fix: register WaitForSubagentsTool in main AgentLoop so it's available in live sessions

Previously wait_for_subagents was only registered inside _run_subagent
(spawned orchestrators). Main conversation agent had no access to it.

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

.github/workflows/build.yml

@@ -0,0 +1,83 @@
+name: Build Nanobot OAuth
+
+on:
+  push:
+    branches: ['main']
+  pull_request:
+    branches: ['main']
+  schedule:
+    - cron: '0 3 * * *'
+  workflow_dispatch:
+
+env:
+  REGISTRY: git.wylab.me
+  IMAGE_NAME: wylab/nanobot
+  BUILDKIT_PROGRESS: plain
+
+jobs:
+  build:
+    runs-on: [self-hosted, linux-amd64]
+    timeout-minutes: 15
+    permissions:
+      contents: read
+      packages: write
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+
+      - name: Log in to the container registry
+        uses: docker/login-action@v3
+        with:
+          registry: ${{ env.REGISTRY }}
+          username: ${{ secrets.REGISTRY_USERNAME || github.actor }}
+          password: ${{ secrets.REGISTRY_PASSWORD || secrets.GITHUB_TOKEN }}
+
+      - name: Build and push Docker image
+        uses: docker/build-push-action@v5
+        with:
+          context: .
+          file: Dockerfile.oauth
+          provenance: false
+          platforms: linux/amd64
+          cache-from: type=registry,ref=${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:buildcache
+          cache-to: type=registry,ref=${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:buildcache,mode=max
+          push: ${{ github.event_name != 'pull_request' }}
+          tags: |
+            ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:latest
+            ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:${{ github.sha }}
+
+  cleanup:
+    if: github.event_name == 'push' || github.event_name == 'schedule'
+    runs-on: [self-hosted, linux-amd64]
+    needs: build
+    steps:
+      - name: Delete images older than 24h
+        env:
+          TOKEN: ${{ secrets.REGISTRY_PASSWORD || secrets.GITHUB_TOKEN }}
+        run: |
+          cutoff=$(date -u -d '24 hours ago' +%s)
+          page=1
+          while true; do
+            versions=$(curl -sf -H "Authorization: token $TOKEN" \
+              "https://${{ env.REGISTRY }}/api/v1/packages/wylab?type=container&q=nanobot&limit=50&page=$page")
+            count=$(echo "$versions" | jq length)
+            [ "$count" = "0" ] && break
+            echo "$versions" | jq -c '.[]' | while read -r pkg; do
+              ver=$(echo "$pkg" | jq -r '.version')
+              # Keep latest and buildcache, only delete SHA tags
+              case "$ver" in latest|buildcache) continue ;; esac
+              created=$(echo "$pkg" | jq -r '.created_at')
+              ts=$(date -u -d "$created" +%s 2>/dev/null || echo 0)
+              if [ "$ts" -lt "$cutoff" ]; then
+                id=$(echo "$pkg" | jq -r '.id')
+                echo "Deleting nanobot:$ver (id=$id, created=$created)"
+                curl -sf -X DELETE -H "Authorization: token $TOKEN" \
+                  "https://${{ env.REGISTRY }}/api/v1/packages/wylab/container/nanobot/$ver" || true
+              fi
+            done
+            [ "$count" -lt 50 ] && break
+            page=$((page + 1))
+          done

Dockerfile.oauth

@@ -0,0 +1,62 @@
+FROM birdxs/nanobot:latest
+
+# ── Skill dependencies ──────────────────────────────────────────────
+
+# APT: ffmpeg (video-frames, whisper), jq, tmux, build-essential (for go)
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    ffmpeg jq tmux build-essential procps && rm -rf /var/lib/apt/lists/*
+
+# gh CLI via GitHub official apt repo
+RUN curl -fsSL https://cli.github.com/packages/githubcli-archive-keyring.gpg \
+    | dd of=/usr/share/keyrings/githubcli-archive-keyring.gpg && \
+    echo "deb [arch=amd64 signed-by=/usr/share/keyrings/githubcli-archive-keyring.gpg] https://cli.github.com/packages stable main" \
+    > /etc/apt/sources.list.d/github-cli.list && \
+    apt-get update && apt-get install -y gh && rm -rf /var/lib/apt/lists/*
+
+# Go toolchain
+RUN curl -fsSL https://go.dev/dl/go1.23.6.linux-amd64.tar.gz | tar -C /usr/local -xzf -
+ENV PATH="/usr/local/go/bin:/root/go/bin:${PATH}"
+
+# Go tools: blogwatcher, blu (blucli), gifgrep, sonos (sonoscli), wacli, songsee
+RUN go install github.com/Hyaxia/blogwatcher/cmd/blogwatcher@latest && \
+    go install github.com/steipete/blucli/cmd/blu@latest && \
+    go install github.com/steipete/gifgrep/cmd/gifgrep@latest && \
+    go install github.com/steipete/sonoscli/cmd/sonos@latest && \
+    go install github.com/steipete/wacli/cmd/wacli@latest && \
+    go install github.com/steipete/songsee/cmd/songsee@latest
+
+# Pre-built binaries from GitHub releases
+# gogcli (gog)
+RUN curl -fsSL https://github.com/steipete/gogcli/releases/download/v0.9.0/gogcli_0.9.0_linux_amd64.tar.gz \
+    | tar -xzf - -C /usr/local/bin gog
+
+# goplaces
+RUN curl -fsSL https://github.com/steipete/goplaces/releases/download/v0.2.1/goplaces_0.2.1_linux_amd64.tar.gz \
+    | tar -xzf - -C /usr/local/bin goplaces
+
+# himalaya (email CLI)
+RUN curl -fsSL https://github.com/pimalaya/himalaya/releases/download/v1.1.0/himalaya.x86_64-linux.tgz \
+    | tar -xzf - -C /usr/local/bin himalaya
+
+# obsidian-cli (release binary is named notesmd-cli, skill expects obsidian-cli)
+RUN curl -fsSL -o /tmp/obsidian.tar.gz https://github.com/yakitrak/obsidian-cli/releases/download/v0.3.0/notesmd-cli_0.3.0_linux_amd64.tar.gz && \
+    tar -xzf /tmp/obsidian.tar.gz -C /tmp notesmd-cli && \
+    mv /tmp/notesmd-cli /usr/local/bin/obsidian-cli && \
+    rm /tmp/obsidian.tar.gz
+
+# Node tools: oracle, gemini-cli, summarize
+RUN npm install -g @steipete/oracle @google/gemini-cli @steipete/summarize
+
+# Python tools: nano-pdf, openai-whisper
+RUN uv tool install nano-pdf && \
+    uv tool install openai-whisper
+ENV PATH="/root/.local/bin:${PATH}"
+
+# ── Nanobot source ──────────────────────────────────────────────────
+
+COPY pyproject.toml README.md LICENSE /app/
+COPY nanobot/ /app/nanobot/
+RUN uv pip install --system --no-cache --reinstall /app psycopg2-binary
+
+ENTRYPOINT ["nanobot"]
+CMD ["gateway"]

nanobot/agent/context.py

@@ -71,7 +71,7 @@ Skills with available="false" need dependencies installed first - you can try in
         return "\n\n---\n\n".join(parts)
     
     def _get_identity(self) -> str:
-        """Get the core identity section."""
+        """Get the core identity section with runtime context."""
         from datetime import datetime
         import time as _time
         now = datetime.now().strftime("%Y-%m-%d %H:%M (%A)")
@@ -79,10 +79,8 @@ Skills with available="false" need dependencies installed first - you can try in
         workspace_path = str(self.workspace.expanduser().resolve())
         system = platform.system()
         runtime = f"{'macOS' if system == 'Darwin' else system} {platform.machine()}, Python {platform.python_version()}"
-        
-        return f"""# nanobot 🐈
 
-You are nanobot, a helpful AI assistant. You have access to tools that allow you to:
+        return f"""You have access to tools that allow you to:
 - Read, write, and edit files
 - Execute shell commands
 - Search the web and fetch web pages

nanobot/agent/loop.py

@@ -2,6 +2,8 @@
 
 import asyncio
 import json
+import time
+from datetime import datetime
 from pathlib import Path
 from typing import Any
 
@@ -17,6 +19,7 @@ from nanobot.agent.tools.shell import ExecTool
 from nanobot.agent.tools.web import WebSearchTool, WebFetchTool
 from nanobot.agent.tools.message import MessageTool
 from nanobot.agent.tools.spawn import SpawnTool
+from nanobot.agent.tools.wait import WaitForSubagentsTool
 from nanobot.agent.tools.cron import CronTool
 from nanobot.agent.memory import MemoryStore
 from nanobot.agent.subagent import SubagentManager
@@ -74,8 +77,10 @@ class AgentLoop:
             exec_config=self.exec_config,
             restrict_to_workspace=restrict_to_workspace,
         )
-        
+
         self._running = False
+        self._quota_cache: dict[str, Any] = {}  # {model: str, cached_at: float}
+        self._quota_cache_ttl: float = 300.0  # 5 minutes
         self._register_default_tools()
     
     def _register_default_tools(self) -> None:
@@ -105,6 +110,7 @@ class AgentLoop:
         # Spawn tool (for subagents)
         spawn_tool = SpawnTool(manager=self.subagents)
         self.tools.register(spawn_tool)
+        self.tools.register(WaitForSubagentsTool(manager=self.subagents))
         
         # Cron tool (for scheduling)
         if self.cron_service:
@@ -143,7 +149,95 @@ class AgentLoop:
         """Stop the agent loop."""
         self._running = False
         logger.info("Agent loop stopping")
-    
+
+    def _select_model_based_on_quota(self) -> str:
+        """Select Opus or Sonnet based on rolling weekly quota burn rate."""
+        # Check cache
+        now = time.time()
+        if self._quota_cache and (now - self._quota_cache.get("cached_at", 0)) < self._quota_cache_ttl:
+            return self._quota_cache["model"]
+
+        # Default models
+        OPUS = "claude-opus-4-6"
+        SONNET = "claude-sonnet-4-6"
+        TOLERANCE = 1.17  # 17% overage triggers downgrade
+
+        # Read rate limits
+        rate_limits_path = self.workspace / "memory" / "rate_limits.json"
+        if not rate_limits_path.exists():
+            logger.warning("rate_limits.json not found, defaulting to Sonnet")
+            return SONNET
+
+        try:
+            with open(rate_limits_path) as f:
+                limits = json.load(f)
+
+            actual_usage = limits.get("weekly_all_models")
+            weekly_reset = limits.get("weekly_reset")
+
+            if actual_usage is None or weekly_reset is None:
+                logger.warning("Rate limit data incomplete, defaulting to Sonnet")
+                return SONNET
+
+            # Calculate expected usage
+            actual_pct = actual_usage * 100
+            week_start = weekly_reset - (168 * 3600)
+            hours_elapsed = max(0, min((now - week_start) / 3600, 168))
+            expected_pct = (hours_elapsed / 168) * 100
+            threshold = expected_pct * TOLERANCE
+
+            # Decision logic
+            if actual_pct > threshold:
+                model = SONNET
+                logger.info(
+                    f"Quota: {actual_pct:.1f}% used, expected {expected_pct:.1f}%, "
+                    f"threshold {threshold:.1f}% → Sonnet"
+                )
+            else:
+                model = OPUS
+                logger.info(
+                    f"Quota: {actual_pct:.1f}% used, expected {expected_pct:.1f}%, "
+                    f"threshold {threshold:.1f}% → Opus"
+                )
+
+            # Cache decision
+            self._quota_cache = {"model": model, "cached_at": now}
+            return model
+
+        except Exception as e:
+            logger.error(f"Error checking quota: {e}, defaulting to Sonnet")
+            return SONNET
+
+    def _get_quota_status(self) -> str:
+        """Return human-readable quota status."""
+        rate_limits_path = self.workspace / "memory" / "rate_limits.json"
+        if not rate_limits_path.exists():
+            return "⚠️ No quota data available yet."
+
+        try:
+            with open(rate_limits_path) as f:
+                limits = json.load(f)
+
+            actual_pct = limits.get("weekly_all_models", 0) * 100
+            reset_ts = limits.get("weekly_reset", 0)
+            now = time.time()
+
+            hours_until_reset = (reset_ts - now) / 3600
+            week_start = reset_ts - (168 * 3600)
+            hours_elapsed = max(0, (now - week_start) / 3600)
+            expected_pct = (hours_elapsed / 168) * 100
+
+            model = self._select_model_based_on_quota()
+
+            return f"""📊 Quota Status:
+• Used: {actual_pct:.1f}% (expected {expected_pct:.1f}%)
+• Resets in: {hours_until_reset:.1f}h
+• Current model: {model}
+• Burn rate: {actual_pct / max(expected_pct, 0.01):.2f}x target"""
+
+        except Exception as e:
+            return f"⚠️ Error reading quota: {e}"
+
     async def _process_message(self, msg: InboundMessage, session_key: str | None = None) -> OutboundMessage | None:
         """
         Process a single inbound message.
@@ -177,8 +271,11 @@ class AgentLoop:
                                   content="🐈 New session started. Memory consolidated.")
         if cmd == "/help":
             return OutboundMessage(channel=msg.channel, chat_id=msg.chat_id,
-                                  content="🐈 nanobot commands:\n/new — Start a new conversation\n/help — Show available commands")
+                                  content="🐈 nanobot commands:\n/new — Start a new conversation\n/help — Show available commands\n/quota — Show quota status")
-        
+        if cmd == "/quota":
+            status = self._get_quota_status()
+            return OutboundMessage(channel=msg.channel, chat_id=msg.chat_id, content=status)
+
         # Consolidate memory before processing if session is too large
         if len(session.messages) > self.memory_window:
             await self._consolidate_memory(session)
@@ -196,28 +293,55 @@ class AgentLoop:
         if isinstance(cron_tool, CronTool):
             cron_tool.set_context(msg.channel, msg.chat_id)
         
+        # Prepend time-gap notice if >5 minutes since last user message
+        current_message = msg.content
+        last_user_ts = None
+        for m in reversed(session.messages):
+            if m.get("role") == "user":
+                last_user_ts = m.get("timestamp")
+                break
+        if last_user_ts:
+            try:
+                last_dt = datetime.fromisoformat(last_user_ts)
+                now_dt = datetime.now()
+                elapsed_seconds = (now_dt - last_dt).total_seconds()
+                if elapsed_seconds > 300:  # 5 minutes
+                    if elapsed_seconds < 3600:
+                        gap_str = f"{int(elapsed_seconds // 60)} minutes"
+                    elif elapsed_seconds < 86400:
+                        gap_str = f"{int(elapsed_seconds // 3600)} hours"
+                    else:
+                        gap_str = f"{int(elapsed_seconds // 86400)} days"
+                    current_message = f"[SYSTEM ANNOUNCEMENT: {gap_str} have elapsed since last user message; take this into account when replying to user]\n\n{msg.content}"
+            except (ValueError, TypeError):
+                pass  # Malformed timestamp — skip silently
+
         # Build initial messages (use get_history for LLM-formatted messages)
         messages = self.context.build_messages(
             history=session.get_history(),
-            current_message=msg.content,
+            current_message=current_message,
             media=msg.media if msg.media else None,
             channel=msg.channel,
             chat_id=msg.chat_id,
         )
-        
+
+        # Select model based on quota
+        selected_model = self._select_model_based_on_quota()
+
         # Agent loop
         iteration = 0
         final_content = None
         tools_used: list[str] = []
-        
+
         while iteration < self.max_iterations:
             iteration += 1
-            
+
             # Call LLM
+            logger.debug(f"Calling LLM with model={selected_model}, provider.thinking_budget={self.provider.thinking_budget}")
             response = await self.provider.chat(
                 messages=messages,
                 tools=self.tools.get_definitions(),
-                model=self.model
+                model=selected_model
             )
             
             # Handle tool calls
@@ -325,14 +449,17 @@ class AgentLoop:
         # Agent loop (limited for announce handling)
         iteration = 0
         final_content = None
-        
+
+        # Select model based on quota
+        selected_model = self._select_model_based_on_quota()
+
         while iteration < self.max_iterations:
             iteration += 1
-            
+
             response = await self.provider.chat(
                 messages=messages,
                 tools=self.tools.get_definitions(),
-                model=self.model
+                model=selected_model
             )
             
             if response.has_tool_calls:
@@ -424,7 +551,9 @@ Respond with ONLY valid JSON, no markdown fences."""
                     {"role": "system", "content": "You are a memory consolidation agent. Respond only with valid JSON."},
                     {"role": "user", "content": prompt},
                 ],
-                model=self.model,
+                model="claude-haiku-4-5",
+                thinking_budget=0,
+                max_tokens=16384,
             )
             text = (response.content or "").strip()
             if text.startswith("```"):

nanobot/agent/skills.py

@@ -170,7 +170,7 @@ class SkillsLoader:
         """Parse nanobot metadata JSON from frontmatter."""
         try:
             data = json.loads(raw)
-            return data.get("nanobot", {}) if isinstance(data, dict) else {}
+            return (data.get("nanobot") or data.get("clawdbot") or {}) if isinstance(data, dict) else {}
         except (json.JSONDecodeError, TypeError):
             return {}
     

nanobot/agent/subagent.py

@@ -15,6 +15,8 @@ from nanobot.agent.tools.registry import ToolRegistry
 from nanobot.agent.tools.filesystem import ReadFileTool, WriteFileTool, EditFileTool, ListDirTool
 from nanobot.agent.tools.shell import ExecTool
 from nanobot.agent.tools.web import WebSearchTool, WebFetchTool
+from nanobot.agent.tools.spawn import SpawnTool
+from nanobot.agent.tools.wait import WaitForSubagentsTool
 
 
 class SubagentManager:
@@ -40,16 +42,21 @@ class SubagentManager:
         self.provider = provider
         self.workspace = workspace
         self.bus = bus
-        self.model = model or provider.get_default_model()
+        # Default to Sonnet, not the provider default (Opus).
+        # Quota switching only affects the main agent's own requests, not SubagentManager.
+        # Explicit model overrides (e.g. Haiku workers) still take precedence.
+        self.model = model or "claude-sonnet-4-6"
         self.brave_api_key = brave_api_key
         self.exec_config = exec_config or ExecToolConfig()
         self.restrict_to_workspace = restrict_to_workspace
         self._running_tasks: dict[str, asyncio.Task[None]] = {}
+        self._task_results: dict[str, str] = {}
     
     async def spawn(
         self,
         task: str,
         label: str | None = None,
+        model: str | None = None,
         origin_channel: str = "cli",
         origin_chat_id: str = "direct",
     ) -> str:
@@ -75,7 +82,7 @@ class SubagentManager:
         
         # Create background task
         bg_task = asyncio.create_task(
-            self._run_subagent(task_id, task, display_label, origin)
+            self._run_subagent(task_id, task, display_label, origin, model=model)
         )
         self._running_tasks[task_id] = bg_task
         
@@ -83,7 +90,7 @@ class SubagentManager:
         bg_task.add_done_callback(lambda _: self._running_tasks.pop(task_id, None))
         
         logger.info(f"Spawned subagent [{task_id}]: {display_label}")
-        return f"Subagent [{display_label}] started (id: {task_id}). I'll notify you when it completes."
+        return f"Subagent [{display_label}] started. Task ID: {task_id}"
     
     async def _run_subagent(
         self,
@@ -91,12 +98,13 @@ class SubagentManager:
         task: str,
         label: str,
         origin: dict[str, str],
+        model: str | None = None,
     ) -> None:
         """Execute the subagent task and announce the result."""
         logger.info(f"Subagent [{task_id}] starting task: {label}")
         
         try:
-            # Build subagent tools (no message tool, no spawn tool)
+            # Build subagent tools (no message tool)
             tools = ToolRegistry()
             allowed_dir = self.workspace if self.restrict_to_workspace else None
             tools.register(ReadFileTool(allowed_dir=allowed_dir))
@@ -110,7 +118,11 @@ class SubagentManager:
             ))
             tools.register(WebSearchTool(api_key=self.brave_api_key))
             tools.register(WebFetchTool())
-            
+            spawn_tool = SpawnTool(manager=self)
+            spawn_tool.set_context("subagent", origin["chat_id"])
+            tools.register(spawn_tool)
+            tools.register(WaitForSubagentsTool(manager=self))
+
             # Build messages with subagent-specific prompt
             system_prompt = self._build_subagent_prompt(task)
             messages: list[dict[str, Any]] = [
@@ -119,7 +131,7 @@ class SubagentManager:
             ]
             
             # Run agent loop (limited iterations)
-            max_iterations = 15
+            max_iterations = 50
             iteration = 0
             final_result: str | None = None
             
@@ -129,7 +141,7 @@ class SubagentManager:
                 response = await self.provider.chat(
                     messages=messages,
                     tools=tools.get_definitions(),
-                    model=self.model,
+                    model=model or self.model,
                 )
                 
                 if response.has_tool_calls:
@@ -188,7 +200,14 @@ class SubagentManager:
     ) -> None:
         """Announce the subagent result to the main agent via the message bus."""
         status_text = "completed successfully" if status == "ok" else "failed"
-        
+
+        # Child subagents (spawned by other subagents) store results silently.
+        # The parent orchestrator collects them via wait_for_subagents.
+        if origin["channel"] == "subagent":
+            self._task_results[task_id] = result
+            logger.debug(f"Subagent [{task_id}] stored result silently (child subagent)")
+            return
+
         announce_content = f"""[Subagent '{label}' {status_text}]
 
 Task: {task}
@@ -197,7 +216,7 @@ Result:
 {result}
 
 Summarize this naturally for the user. Keep it brief (1-2 sentences). Do not mention technical details like "subagent" or task IDs."""
-        
+
         # Inject as system message to trigger main agent
         msg = InboundMessage(
             channel="system",
@@ -205,7 +224,7 @@ Summarize this naturally for the user. Keep it brief (1-2 sentences). Do not men
             chat_id=f"{origin['channel']}:{origin['chat_id']}",
             content=announce_content,
         )
-        
+
         await self.bus.publish_inbound(msg)
         logger.debug(f"Subagent [{task_id}] announced result to {origin['channel']}:{origin['chat_id']}")
     
@@ -237,7 +256,6 @@ You are a subagent spawned by the main agent to complete a specific task.
 
 ## What You Cannot Do
 - Send messages directly to users (no message tool available)
-- Spawn other subagents
 - Access the main agent's conversation history
 
 ## Workspace
@@ -246,6 +264,25 @@ Skills are available at: {self.workspace}/skills/ (read SKILL.md files as needed
 
 When you have completed the task, provide a clear summary of your findings or actions."""
     
+    async def wait_for(self, task_ids: list[str]) -> str:
+        """Wait for specified child subagents to complete and return their results."""
+        tasks_to_wait = [
+            self._running_tasks[tid]
+            for tid in task_ids
+            if tid in self._running_tasks
+        ]
+        if tasks_to_wait:
+            await asyncio.gather(*tasks_to_wait, return_exceptions=True)
+
+        results = []
+        for tid in task_ids:
+            result = self._task_results.get(tid)
+            if result is not None:
+                results.append(f"[{tid}]:\n{result}")
+            else:
+                results.append(f"[{tid}]: No result found (invalid ID or task failed before storing)")
+        return "\n\n---\n\n".join(results)
+
     def get_running_count(self) -> int:
         """Return the number of currently running subagents."""
         return len(self._running_tasks)

nanobot/agent/tools/spawn.py

@@ -51,15 +51,20 @@ class SpawnTool(Tool):
                     "type": "string",
                     "description": "Optional short label for the task (for display)",
                 },
+                "model": {
+                    "type": "string",
+                    "description": "Optional model override for the subagent (e.g. 'claude-haiku-4-5'). Defaults to the main agent's model.",
+                },
             },
             "required": ["task"],
         }
     
-    async def execute(self, task: str, label: str | None = None, **kwargs: Any) -> str:
+    async def execute(self, task: str, label: str | None = None, model: str | None = None, **kwargs: Any) -> str:
         """Spawn a subagent to execute the given task."""
         return await self._manager.spawn(
             task=task,
             label=label,
+            model=model,
             origin_channel=self._origin_channel,
             origin_chat_id=self._origin_chat_id,
         )

nanobot/agent/tools/wait.py

@@ -0,0 +1,50 @@
+"""Wait-for-subagents tool for orchestrator subagents."""
+
+from typing import Any, TYPE_CHECKING
+
+from nanobot.agent.tools.base import Tool
+
+if TYPE_CHECKING:
+    from nanobot.agent.subagent import SubagentManager
+
+
+class WaitForSubagentsTool(Tool):
+    """
+    Tool to wait for child subagents to complete and collect their results.
+
+    Use this after spawning multiple subagents to wait for all of them
+    and get their results for synthesis.
+    """
+
+    def __init__(self, manager: "SubagentManager"):
+        self._manager = manager
+
+    @property
+    def name(self) -> str:
+        return "wait_for_subagents"
+
+    @property
+    def description(self) -> str:
+        return (
+            "Wait for one or more child subagents to complete and return their results. "
+            "Use this after spawning subagents to collect all results before synthesizing. "
+            "Blocks until all specified subagents finish."
+        )
+
+    @property
+    def parameters(self) -> dict[str, Any]:
+        return {
+            "type": "object",
+            "properties": {
+                "task_ids": {
+                    "type": "array",
+                    "items": {"type": "string"},
+                    "description": "List of task IDs to wait for (from spawn tool responses)",
+                },
+            },
+            "required": ["task_ids"],
+        }
+
+    async def execute(self, task_ids: list[str], **kwargs: Any) -> str:
+        """Wait for the specified subagents and return their results."""
+        return await self._manager.wait_for(task_ids)

nanobot/channels/base.py

@@ -69,11 +69,15 @@ class BaseChannel(ABC):
             True if allowed, False otherwise.
         """
         allow_list = getattr(self.config, "allow_from", [])
-        
+
         # If no allow list, allow everyone
         if not allow_list:
             return True
-        
+
+        # Wildcard allows everyone
+        if "*" in allow_list:
+            return True
+
         sender_str = str(sender_id)
         if sender_str in allow_list:
             return True

nanobot/channels/telegram.py

@@ -92,6 +92,7 @@ class TelegramChannel(BaseChannel):
         BotCommand("start", "Start the bot"),
         BotCommand("new", "Start a new conversation"),
         BotCommand("help", "Show available commands"),
+        BotCommand("quota", "Show current quota status"),
     ]
     
     def __init__(
@@ -127,6 +128,7 @@ class TelegramChannel(BaseChannel):
         self._app.add_handler(CommandHandler("start", self._on_start))
         self._app.add_handler(CommandHandler("new", self._forward_command))
         self._app.add_handler(CommandHandler("help", self._forward_command))
+        self._app.add_handler(CommandHandler("quota", self._forward_command))
         
         # Add message handler for text, photos, voice, documents
         self._app.add_handler(

nanobot/cli/commands.py

@@ -270,20 +270,21 @@ This file stores important information that should persist across sessions.
 
 
 def _make_provider(config):
-    """Create LiteLLMProvider from config. Exits if no API key found."""
+    """Create LLM provider from config. Uses OAuth for subscription tokens."""
-    from nanobot.providers.litellm_provider import LiteLLMProvider
+    from nanobot.providers import create_provider
     p = config.get_provider()
     model = config.agents.defaults.model
     if not (p and p.api_key) and not model.startswith("bedrock/"):
         console.print("[red]Error: No API key configured.[/red]")
         console.print("Set one in ~/.nanobot/config.json under providers section")
         raise typer.Exit(1)
-    return LiteLLMProvider(
+    return create_provider(
         api_key=p.api_key if p else None,
+        model=model,
         api_base=config.get_api_base(),
-        default_model=model,
         extra_headers=p.extra_headers if p else None,
         provider_name=config.get_provider_name(),
+        thinking_budget=config.agents.defaults.thinking_budget,
     )
 
 
@@ -507,6 +508,9 @@ def agent(
 channels_app = typer.Typer(help="Manage channels")
 app.add_typer(channels_app, name="channels")
 
+from nanobot.cli.oauth import oauth_app
+app.add_typer(oauth_app, name="oauth")
+
 
 @channels_app.command("status")
 def channels_status():

nanobot/cli/oauth.py

@@ -0,0 +1,93 @@
+"""OAuth CLI commands for subscription authentication."""
+
+from pathlib import Path
+from typing import Optional
+
+import typer
+from rich.console import Console
+
+oauth_app = typer.Typer(help="Manage OAuth authentication for subscription-based providers")
+console = Console()
+
+
+@oauth_app.command("login")
+def login(
+    provider: str = typer.Argument("anthropic", help="Provider name"),
+    token: Optional[str] = typer.Option(None, "--token", "-t", help="OAuth token (from claude setup-token)"),
+):
+    """Login to a provider using OAuth.
+
+    For Anthropic Claude Max/Pro, run 'claude setup-token' and paste the token here.
+
+    Example:
+        nanobot oauth login anthropic --token sk-ant-oat01-xxx
+    """
+    from nanobot.config.oauth_store import OAuthStore
+    from nanobot.config.schema import OAuthCredentials
+
+    if provider != "anthropic":
+        console.print(f"[red]OAuth login for {provider} not yet supported[/red]")
+        return
+
+    if not token:
+        console.print("Please provide your OAuth token:")
+        console.print("  1. Run: claude setup-token")
+        console.print("  2. Copy the sk-ant-oat01-... token")
+        console.print("  3. Run: nanobot oauth login anthropic --token <your-token>")
+        console.print()
+        token = typer.prompt("Token", hide_input=True)
+
+    if not token or "sk-ant-oat" not in token:
+        console.print("[red]Invalid token. Must contain sk-ant-oat[/red]")
+        return
+
+    store = OAuthStore(Path.home() / ".nanobot")
+    creds = OAuthCredentials(
+        access_token=token,
+        token_type="token"  # setup-token doesn't expire
+    )
+    store.save(provider, creds)
+
+    console.print(f"[green]Successfully saved {provider} OAuth credentials![/green]")
+
+
+@oauth_app.command("status")
+def status():
+    """Show OAuth credential status."""
+    from nanobot.config.oauth_store import OAuthStore
+
+    store = OAuthStore(Path.home() / ".nanobot")
+
+    providers = ["anthropic"]
+    found_any = False
+
+    for provider in providers:
+        creds = store.load(provider)
+        if creds:
+            found_any = True
+            st = "valid"
+            if creds.is_expired:
+                st = "EXPIRED"
+            elif creds.expires_soon:
+                st = "expires soon"
+
+            token_preview = creds.access_token[:20] + "..."
+            console.print(f"  {provider}: {token_preview} ({st})")
+
+    if not found_any:
+        console.print("No OAuth credentials configured.")
+        console.print("Run: nanobot oauth login anthropic --token <token>")
+
+
+@oauth_app.command("logout")
+def logout(
+    provider: str = typer.Argument("anthropic", help="Provider name"),
+):
+    """Remove OAuth credentials for a provider."""
+    from nanobot.config.oauth_store import OAuthStore
+
+    store = OAuthStore(Path.home() / ".nanobot")
+    if store.delete(provider):
+        console.print(f"[green]Removed {provider} OAuth credentials[/green]")
+    else:
+        console.print(f"No credentials found for {provider}")

nanobot/config/loader.py

@@ -12,35 +12,53 @@ def get_config_path() -> Path:
     return Path.home() / ".nanobot" / "config.json"
 
 
+def _get_oauth_store_dir() -> Path:
+    """Get the OAuth store directory."""
+    return Path.home() / ".nanobot"
+
+
 def get_data_dir() -> Path:
     """Get the nanobot data directory."""
     from nanobot.utils.helpers import get_data_path
     return get_data_path()
 
 
+def _inject_oauth_credentials(config: Config) -> Config:
+    """Inject OAuth credentials from store into config if available."""
+    from nanobot.config.oauth_store import OAuthStore
+
+    store = OAuthStore(_get_oauth_store_dir())
+    creds = store.load("anthropic")
+    if creds and creds.access_token and not creds.is_expired:
+        config.providers.anthropic.api_key = creds.access_token
+
+    return config
+
+
 def load_config(config_path: Path | None = None) -> Config:
     """
     Load configuration from file or create default.
-    
+
     Args:
         config_path: Optional path to config file. Uses default if not provided.
-    
+
     Returns:
         Loaded configuration object.
     """
     path = config_path or get_config_path()
-    
+
     if path.exists():
         try:
             with open(path) as f:
                 data = json.load(f)
             data = _migrate_config(data)
-            return Config.model_validate(convert_keys(data))
+            config = Config.model_validate(convert_keys(data))
+            return _inject_oauth_credentials(config)
         except (json.JSONDecodeError, ValueError) as e:
             print(f"Warning: Failed to load config from {path}: {e}")
             print("Using default configuration.")
-    
+
-    return Config()
+    return _inject_oauth_credentials(Config())
 
 
 def save_config(config: Config, config_path: Path | None = None) -> None:

nanobot/config/oauth_store.py

@@ -0,0 +1,59 @@
+"""OAuth credential storage."""
+
+import json
+from pathlib import Path
+from typing import Any
+
+from nanobot.config.schema import OAuthCredentials
+
+
+class OAuthStore:
+    """Stores OAuth credentials in a JSON file."""
+
+    FILENAME = "oauth-credentials.json"
+
+    def __init__(self, config_dir: Path):
+        self.config_dir = config_dir
+        self.file_path = config_dir / self.FILENAME
+
+    def _load_all(self) -> dict[str, Any]:
+        """Load all credentials from file."""
+        if not self.file_path.exists():
+            return {}
+
+        with open(self.file_path, "r") as f:
+            return json.load(f)
+
+    def _save_all(self, data: dict[str, Any]) -> None:
+        """Save all credentials to file."""
+        self.config_dir.mkdir(parents=True, exist_ok=True)
+
+        with open(self.file_path, "w") as f:
+            json.dump(data, f, indent=2)
+
+        # Secure permissions
+        self.file_path.chmod(0o600)
+
+    def save(self, provider: str, credentials: OAuthCredentials) -> None:
+        """Save credentials for a provider."""
+        data = self._load_all()
+        data[provider] = credentials.model_dump()
+        self._save_all(data)
+
+    def load(self, provider: str) -> OAuthCredentials | None:
+        """Load credentials for a provider."""
+        data = self._load_all()
+        if provider not in data:
+            return None
+
+        return OAuthCredentials(**data[provider])
+
+    def delete(self, provider: str) -> bool:
+        """Delete credentials for a provider."""
+        data = self._load_all()
+        if provider not in data:
+            return False
+
+        del data[provider]
+        self._save_all(data)
+        return True

nanobot/config/schema.py

@@ -163,6 +163,7 @@ class AgentDefaults(BaseModel):
     temperature: float = 0.7
     max_tool_iterations: int = 20
     memory_window: int = 50
+    thinking_budget: int = 0  # 0 = disabled; >0 = token budget for extended thinking
 
 
 class AgentsConfig(BaseModel):
@@ -170,11 +171,41 @@ class AgentsConfig(BaseModel):
     defaults: AgentDefaults = Field(default_factory=AgentDefaults)
 
 
+class OAuthCredentials(BaseModel):
+    """OAuth token credentials for subscription-based auth."""
+    access_token: str = ""
+    refresh_token: str = ""
+    expires_at: int = 0  # Unix timestamp
+    token_type: str = "oauth"  # "oauth" or "token" (setup-token)
+
+    @property
+    def is_oauth_token(self) -> bool:
+        """Check if this is an OAuth token (vs regular API key)."""
+        return "sk-ant-oat" in self.access_token
+
+    @property
+    def is_expired(self) -> bool:
+        """Check if token has expired."""
+        import time
+        if self.expires_at == 0:
+            return False  # No expiry set (setup-token)
+        return time.time() > self.expires_at
+
+    @property
+    def expires_soon(self) -> bool:
+        """Check if token expires within 10 minutes."""
+        import time
+        if self.expires_at == 0:
+            return False
+        return time.time() > (self.expires_at - 600)
+
+
 class ProviderConfig(BaseModel):
     """LLM provider configuration."""
     api_key: str = ""
     api_base: str | None = None
     extra_headers: dict[str, str] | None = None  # Custom headers (e.g. APP-Code for AiHubMix)
+    oauth_credentials: OAuthCredentials | None = None
 
 
 class ProvidersConfig(BaseModel):

nanobot/providers/__init__.py

@@ -1,6 +1,45 @@
-"""LLM provider abstraction module."""
+"""Provider module exports."""
 
-from nanobot.providers.base import LLMProvider, LLMResponse
+from nanobot.providers.base import LLMProvider, LLMResponse, ToolCallRequest
 from nanobot.providers.litellm_provider import LiteLLMProvider
+from nanobot.providers.anthropic_oauth import AnthropicOAuthProvider
+from nanobot.providers.registry import should_use_oauth_provider
 
-__all__ = ["LLMProvider", "LLMResponse", "LiteLLMProvider"]
+__all__ = [
+    "LLMProvider",
+    "LLMResponse",
+    "ToolCallRequest",
+    "LiteLLMProvider",
+    "AnthropicOAuthProvider",
+    "create_provider",
+]
+
+
+def create_provider(
+    api_key: str,
+    model: str,
+    api_base: str | None = None,
+    extra_headers: dict[str, str] | None = None,
+    provider_name: str | None = None,
+    thinking_budget: int = 0,
+) -> LLMProvider:
+    """Factory function to create appropriate provider.
+
+    Automatically selects AnthropicOAuthProvider for OAuth tokens,
+    LiteLLMProvider for everything else.
+    """
+    if should_use_oauth_provider(api_key, model):
+        return AnthropicOAuthProvider(
+            oauth_token=api_key,
+            default_model=model,
+            api_base=api_base,
+            thinking_budget=thinking_budget,
+        )
+
+    return LiteLLMProvider(
+        api_key=api_key,
+        api_base=api_base,
+        default_model=model,
+        extra_headers=extra_headers,
+        provider_name=provider_name,
+    )

nanobot/providers/anthropic_oauth.py

@@ -0,0 +1,410 @@
+"""Anthropic OAuth provider - direct API calls with Bearer auth.
+
+This provider bypasses litellm to properly handle OAuth tokens
+which require Authorization: Bearer header instead of x-api-key.
+"""
+
+import json
+from typing import Any
+
+import httpx
+from loguru import logger
+
+from nanobot.providers.base import LLMProvider, LLMResponse, ToolCallRequest
+from nanobot.providers.oauth_utils import get_auth_headers
+
+
+class AnthropicOAuthProvider(LLMProvider):
+    """
+    Anthropic provider using OAuth token authentication.
+
+    Unlike the LiteLLM provider, this calls the Anthropic API directly
+    with proper Bearer token authentication for Claude Max/Pro subscriptions.
+    """
+
+    ANTHROPIC_API_URL = "https://api.anthropic.com/v1/messages"
+
+    def __init__(
+        self,
+        oauth_token: str,
+        default_model: str = "claude-opus-4-5",
+        api_base: str | None = None,
+        thinking_budget: int = 0,
+    ):
+        super().__init__(api_key=None, api_base=api_base)
+        self.oauth_token = oauth_token
+        self.default_model = default_model
+        self.thinking_budget = thinking_budget
+        self._client: httpx.AsyncClient | None = None
+
+    def _get_headers(self) -> dict[str, str]:
+        """Get request headers with Bearer auth."""
+        return get_auth_headers(self.oauth_token, is_oauth=True)
+
+    def _get_api_url(self) -> str:
+        """Get API endpoint URL."""
+        if self.api_base:
+            return f"{self.api_base.rstrip('/')}/v1/messages"
+        return self.ANTHROPIC_API_URL
+
+    @staticmethod
+    def _normalize_model(model: str) -> str:
+        """Normalize model name for the Anthropic API.
+
+        Anthropic model IDs use hyphens (claude-sonnet-4-5), but users often
+        write dots (claude-sonnet-4.5). Normalize so both work.
+        """
+        return model.replace(".", "-")
+
+    async def _get_client(self) -> httpx.AsyncClient:
+        """Get or create async HTTP client."""
+        if self._client is None:
+            self._client = httpx.AsyncClient(timeout=300.0)
+        return self._client
+
+    def _prepare_messages(
+        self,
+        messages: list[dict[str, Any]]
+    ) -> tuple[str | None, list[dict[str, Any]]]:
+        """Prepare messages: extract system prompt and convert OpenAI format to Anthropic.
+
+        The agent loop produces messages in OpenAI format:
+          - assistant msgs with tool_calls [{type:"function", function:{name, arguments}}]
+          - tool role msgs with tool_call_id, name, content
+
+        Anthropic API expects:
+          - assistant msgs with content blocks [{type:"tool_use", id, name, input}]
+          - user msgs with content blocks [{type:"tool_result", tool_use_id, content}]
+
+        Returns (system_prompt, anthropic_messages)
+        """
+        system_parts = []
+        converted: list[dict[str, Any]] = []
+
+        for msg in messages:
+            role = msg.get("role")
+
+            if role == "system":
+                system_parts.append(msg.get("content", ""))
+                continue
+
+            if role == "assistant" and msg.get("tool_calls"):
+                # Convert OpenAI tool_calls to Anthropic content blocks
+                content_blocks: list[dict[str, Any]] = []
+                # Preserve thinking blocks (list=raw API blocks with signatures, str=legacy)
+                rc = msg.get("reasoning_content")
+                if isinstance(rc, list):
+                    content_blocks.extend(rc)
+                elif isinstance(rc, str) and rc:
+                    content_blocks.append({"type": "thinking", "thinking": rc})
+                text = msg.get("content")
+                if text:
+                    content_blocks.append({"type": "text", "text": text})
+                for tc in msg["tool_calls"]:
+                    func = tc.get("function", {})
+                    args = func.get("arguments", "{}")
+                    if isinstance(args, str):
+                        try:
+                            args = json.loads(args)
+                        except (json.JSONDecodeError, TypeError):
+                            args = {}
+                    content_blocks.append({
+                        "type": "tool_use",
+                        "id": tc.get("id", ""),
+                        "name": func.get("name", ""),
+                        "input": args,
+                    })
+                converted.append({"role": "assistant", "content": content_blocks})
+                continue
+
+            if role == "assistant" and msg.get("reasoning_content"):
+                # Plain assistant message with thinking (no tool calls)
+                rc = msg["reasoning_content"]
+                if isinstance(rc, list):
+                    content_blocks = list(rc)
+                else:
+                    content_blocks = [{"type": "thinking", "thinking": rc}]
+                text = msg.get("content")
+                if text:
+                    content_blocks.append({"type": "text", "text": text})
+                converted.append({"role": "assistant", "content": content_blocks})
+                continue
+
+            if role == "tool":
+                # Convert tool result to Anthropic user message with tool_result block
+                tool_result_block = {
+                    "type": "tool_result",
+                    "tool_use_id": msg.get("tool_call_id", ""),
+                    "content": msg.get("content", ""),
+                }
+                # Merge into previous user message if it already has tool_result blocks
+                if converted and converted[-1].get("role") == "user":
+                    prev_content = converted[-1].get("content")
+                    if isinstance(prev_content, list):
+                        prev_content.append(tool_result_block)
+                        continue
+                converted.append({"role": "user", "content": [tool_result_block]})
+                continue
+
+            if role == "user":
+                content = msg.get("content", "")
+                # Convert OpenAI image_url blocks to Anthropic image blocks
+                if isinstance(content, list):
+                    content = self._convert_image_blocks(content)
+                # Merge text into previous user message if it has tool_result blocks
+                # (handles the "Reflect on the results" interleaved message)
+                if converted and converted[-1].get("role") == "user":
+                    prev_content = converted[-1].get("content")
+                    if isinstance(prev_content, list):
+                        if isinstance(content, str):
+                            prev_content.append({"type": "text", "text": content})
+                        elif isinstance(content, list):
+                            prev_content.extend(content)
+                        continue
+                converted.append({"role": role, "content": content})
+                continue
+
+            # Pass through other messages (assistant without tool_calls, etc.)
+            converted.append(msg)
+
+        system_prompt = "\n\n".join(system_parts)
+        return system_prompt, converted
+
+    @staticmethod
+    def _convert_image_blocks(content: list[dict[str, Any]]) -> list[dict[str, Any]]:
+        """Convert OpenAI image_url blocks to Anthropic image blocks.
+
+        OpenAI format:  {"type": "image_url", "image_url": {"url": "data:mime;base64,DATA"}}
+        Anthropic format: {"type": "image", "source": {"type": "base64", "media_type": "mime", "data": "DATA"}}
+        """
+        converted = []
+        for block in content:
+            if block.get("type") == "image_url":
+                url = block.get("image_url", {}).get("url", "")
+                if url.startswith("data:") and ";base64," in url:
+                    header, data = url.split(";base64,", 1)
+                    media_type = header.removeprefix("data:")
+                    converted.append({
+                        "type": "image",
+                        "source": {"type": "base64", "media_type": media_type, "data": data},
+                    })
+                else:
+                    converted.append({
+                        "type": "image",
+                        "source": {"type": "url", "url": url},
+                    })
+            else:
+                converted.append(block)
+        return converted
+
+    def _convert_tools_to_anthropic(
+        self,
+        tools: list[dict[str, Any]] | None
+    ) -> list[dict[str, Any]] | None:
+        """Convert OpenAI-format tools to Anthropic format."""
+        if not tools:
+            return None
+
+        anthropic_tools = []
+        for tool in tools:
+            if tool.get("type") == "function":
+                func = tool["function"]
+                anthropic_tools.append({
+                    "name": func["name"],
+                    "description": func.get("description", ""),
+                    "input_schema": func.get("parameters", {"type": "object", "properties": {}})
+                })
+
+        return anthropic_tools if anthropic_tools else None
+
+    async def _make_request(
+        self,
+        messages: list[dict[str, Any]],
+        system: str | None = None,
+        model: str = "claude-opus-4-5",
+        max_tokens: int = 4096,
+        temperature: float = 0.7,
+        tools: list[dict[str, Any]] | None = None,
+        thinking_budget_override: int | None = None,
+    ) -> dict[str, Any]:
+        """Make request to Anthropic API."""
+        client = await self._get_client()
+
+        payload: dict[str, Any] = {
+            "model": model,
+            "messages": messages,
+            "max_tokens": max_tokens,
+        }
+
+        # Extended thinking: temperature must be 1 when enabled
+        effective_thinking = thinking_budget_override if thinking_budget_override is not None else self.thinking_budget
+        if effective_thinking > 0:
+            payload["temperature"] = 1
+            # max_tokens must exceed budget_tokens
+            if max_tokens <= effective_thinking:
+                payload["max_tokens"] = effective_thinking + 4096
+            payload["thinking"] = {
+                "type": "enabled",
+                "budget_tokens": effective_thinking,
+            }
+        else:
+            payload["temperature"] = temperature
+
+        if system:
+            payload["system"] = system
+
+        if tools:
+            payload["tools"] = tools
+
+        logger.info(
+            "Anthropic request: model={} max_tokens={} thinking={} tools={}",
+            payload.get("model"), payload.get("max_tokens"),
+            payload.get("thinking", "disabled"),
+            len(payload.get("tools", [])),
+        )
+
+        response = await client.post(
+            self._get_api_url(),
+            headers=self._get_headers(),
+            json=payload,
+        )
+
+        # Dump rate limit headers for analysis
+        try:
+            import datetime
+            import os
+            header_dump = {
+                "timestamp": datetime.datetime.utcnow().isoformat(),
+                "status_code": response.status_code,
+                "model": payload.get("model"),
+                "headers": dict(response.headers),
+            }
+            dump_path = "/root/.nanobot/workspace/api_headers.jsonl"
+            with open(dump_path, "a") as f:
+                f.write(json.dumps(header_dump) + "\n")
+            # Capture rate limit state for quota-based model switching
+            hdrs = response.headers
+            rate_limit_state = {
+                "updated_at": datetime.datetime.utcnow().isoformat(),
+                "model": payload.get("model"),
+                "weekly_all_models": float(hdrs["anthropic-ratelimit-unified-7d-utilization"]) if hdrs.get("anthropic-ratelimit-unified-7d-utilization") else None,
+                "weekly_sonnet": float(hdrs["anthropic-ratelimit-unified-7d_sonnet-utilization"]) if hdrs.get("anthropic-ratelimit-unified-7d_sonnet-utilization") else None,
+                "session_5h": float(hdrs["anthropic-ratelimit-unified-5h-utilization"]) if hdrs.get("anthropic-ratelimit-unified-5h-utilization") else None,
+                "weekly_reset": int(hdrs["anthropic-ratelimit-unified-7d-reset"]) if hdrs.get("anthropic-ratelimit-unified-7d-reset") else None,
+                "session_reset": int(hdrs["anthropic-ratelimit-unified-5h-reset"]) if hdrs.get("anthropic-ratelimit-unified-5h-reset") else None,
+                "binding_limit": hdrs.get("anthropic-ratelimit-unified-representative-claim"),
+                "sonnet_fallback": hdrs.get("anthropic-ratelimit-unified-fallback"),
+            }
+            state_path = "/root/.nanobot/workspace/memory/rate_limits.json"
+            os.makedirs(os.path.dirname(state_path), exist_ok=True)
+            with open(state_path, "w") as f:
+                json.dump(rate_limit_state, f, indent=2)
+        except Exception as e:
+            logger.warning("Rate limit header capture failed: {}", e)
+
+        if response.status_code != 200:
+            error_text = response.text
+            raise Exception(f"Anthropic API error {response.status_code}: {error_text}")
+
+        return response.json()
+
+    async def chat(
+        self,
+        messages: list[dict[str, Any]],
+        tools: list[dict[str, Any]] | None = None,
+        model: str | None = None,
+        max_tokens: int = 4096,
+        temperature: float = 0.7,
+        thinking_budget: int | None = None,
+    ) -> LLMResponse:
+        """Send chat completion request to Anthropic API."""
+        model = model or self.default_model
+
+        # Strip provider prefix if present (e.g. "anthropic/claude-opus-4-5" -> "claude-opus-4-5")
+        if "/" in model:
+            model = model.split("/")[-1]
+
+        # Normalize dots to hyphens (claude-sonnet-4.5 -> claude-sonnet-4-5)
+        model = self._normalize_model(model)
+
+        system, prepared_messages = self._prepare_messages(messages)
+        anthropic_tools = self._convert_tools_to_anthropic(tools)
+
+        # Per-call thinking override (None = use instance default)
+        effective_thinking = self.thinking_budget if thinking_budget is None else thinking_budget
+
+        try:
+            response = await self._make_request(
+                messages=prepared_messages,
+                system=system,
+                model=model,
+                max_tokens=max_tokens,
+                temperature=temperature,
+                tools=anthropic_tools,
+                thinking_budget_override=effective_thinking,
+            )
+            return self._parse_response(response)
+        except Exception as e:
+            return LLMResponse(
+                content=f"Error calling LLM: {str(e)}",
+                finish_reason="error",
+            )
+
+    def _parse_response(self, response: dict[str, Any]) -> LLMResponse:
+        """Parse Anthropic API response."""
+        content_blocks = response.get("content", [])
+
+        text_content = ""
+        thinking_blocks: list[dict[str, Any]] = []
+        tool_calls = []
+
+        for block in content_blocks:
+            if block.get("type") == "thinking":
+                # Preserve full block including signature for multi-turn replay
+                thinking_blocks.append(block)
+            elif block.get("type") == "text":
+                text_content += block.get("text", "")
+            elif block.get("type") == "tool_use":
+                tool_calls.append(ToolCallRequest(
+                    id=block.get("id", ""),
+                    name=block.get("name", ""),
+                    arguments=block.get("input", {}),
+                ))
+
+        usage = {}
+        if "usage" in response:
+            usage = {
+                "prompt_tokens": response["usage"].get("input_tokens", 0),
+                "completion_tokens": response["usage"].get("output_tokens", 0),
+                "total_tokens": (
+                    response["usage"].get("input_tokens", 0) +
+                    response["usage"].get("output_tokens", 0)
+                ),
+            }
+
+        stop_reason = response.get("stop_reason", "end_turn")
+        thinking_chars = sum(len(b.get("thinking", "")) for b in thinking_blocks) if thinking_blocks else 0
+        logger.info(
+            "Anthropic response: stop={} tool_calls={} thinking={} chars, "
+            "input={} output={} tokens",
+            stop_reason, len(tool_calls), thinking_chars,
+            usage.get("prompt_tokens", 0), usage.get("completion_tokens", 0),
+        )
+
+        return LLMResponse(
+            content=text_content or None,
+            tool_calls=tool_calls,
+            finish_reason=stop_reason,
+            usage=usage,
+            reasoning_content=thinking_blocks or None,
+        )
+
+    def get_default_model(self) -> str:
+        """Get the default model."""
+        return self.default_model
+
+    async def close(self):
+        """Close the HTTP client."""
+        if self._client:
+            await self._client.aclose()
+            self._client = None

nanobot/providers/base.py

@@ -20,7 +20,7 @@ class LLMResponse:
     tool_calls: list[ToolCallRequest] = field(default_factory=list)
     finish_reason: str = "stop"
     usage: dict[str, int] = field(default_factory=dict)
-    reasoning_content: str | None = None  # Kimi, DeepSeek-R1 etc.
+    reasoning_content: Any = None  # str for Kimi/DeepSeek-R1; list[dict] for Anthropic thinking blocks
     
     @property
     def has_tool_calls(self) -> bool:
@@ -48,6 +48,7 @@ class LLMProvider(ABC):
         model: str | None = None,
         max_tokens: int = 4096,
         temperature: float = 0.7,
+        thinking_budget: int | None = None,
     ) -> LLMResponse:
         """
         Send a chat completion request.

nanobot/providers/litellm_provider.py

@@ -106,6 +106,7 @@ class LiteLLMProvider(LLMProvider):
         model: str | None = None,
         max_tokens: int = 4096,
         temperature: float = 0.7,
+        thinking_budget: int | None = None,
     ) -> LLMResponse:
         """
         Send a chat completion request via LiteLLM.

nanobot/providers/oauth_utils.py

@@ -0,0 +1,38 @@
+"""OAuth utility functions for Anthropic subscription auth."""
+
+from typing import Any
+
+
+def is_oauth_token(token: str | None) -> bool:
+    """Check if token is an OAuth token (vs regular API key).
+
+    OAuth tokens from Claude Max/Pro contain 'sk-ant-oat' prefix.
+    Regular API keys use 'sk-ant-api03' or similar.
+    """
+    if not token:
+        return False
+    return "sk-ant-oat" in token
+
+
+def get_auth_headers(token: str, is_oauth: bool = False) -> dict[str, str]:
+    """Get authentication headers for Anthropic API.
+
+    OAuth tokens require Authorization: Bearer header.
+    Regular API keys use x-api-key header.
+    """
+    headers: dict[str, str] = {
+        "anthropic-version": "2023-06-01",
+        "content-type": "application/json",
+    }
+
+    if is_oauth:
+        headers["Authorization"] = f"Bearer {token}"
+        # Required headers to mimic Claude Code client
+        headers["anthropic-beta"] = "claude-code-20250219,oauth-2025-04-20"
+        headers["anthropic-dangerous-direct-browser-access"] = "true"
+        headers["user-agent"] = "claude-cli/2.1.2 (external, cli)"
+        headers["x-app"] = "cli"
+    else:
+        headers["x-api-key"] = token
+
+    return headers

nanobot/providers/registry.py

@@ -357,3 +357,24 @@ def find_by_name(name: str) -> ProviderSpec | None:
         if spec.name == name:
             return spec
     return None
+
+
+def should_use_oauth_provider(api_key: str | None, model: str) -> bool:
+    """Determine if OAuth provider should be used.
+
+    OAuth provider is used when:
+    1. API key is an OAuth token (contains 'sk-ant-oat')
+    2. Model is an Anthropic model (contains 'claude' or 'anthropic')
+    """
+    if not api_key:
+        return False
+
+    if "sk-ant-oat" not in api_key:
+        return False
+
+    model_lower = model.lower()
+    anthropic_spec = find_by_name("anthropic")
+    if anthropic_spec:
+        return any(kw in model_lower for kw in anthropic_spec.keywords)
+
+    return False

tests/test_anthropic_oauth.py

@@ -0,0 +1,88 @@
+"""Test Anthropic OAuth provider."""
+import pytest
+from unittest.mock import AsyncMock, patch, MagicMock
+from nanobot.providers.anthropic_oauth import AnthropicOAuthProvider
+from nanobot.providers.base import LLMResponse
+
+
+@pytest.fixture
+def provider():
+    """Create provider with test OAuth token."""
+    return AnthropicOAuthProvider(
+        oauth_token="sk-ant-oat01-test-token",
+        default_model="claude-opus-4-5"
+    )
+
+
+def test_provider_init(provider):
+    """Provider should initialize with OAuth token."""
+    assert provider.oauth_token == "sk-ant-oat01-test-token"
+    assert provider.default_model == "claude-opus-4-5"
+
+
+def test_provider_uses_bearer_auth(provider):
+    """Provider should use Bearer auth, not x-api-key."""
+    headers = provider._get_headers()
+    assert "Authorization" in headers
+    assert headers["Authorization"].startswith("Bearer ")
+    assert "x-api-key" not in headers
+
+
+@pytest.mark.asyncio
+async def test_chat_prepends_system_prompt(provider):
+    """Chat should prepend Claude Code identity to system prompt."""
+    messages = [{"role": "user", "content": "Hello"}]
+
+    with patch.object(provider, "_make_request", new_callable=AsyncMock) as mock:
+        mock.return_value = {"content": [{"type": "text", "text": "Hi"}], "stop_reason": "end_turn"}
+        await provider.chat(messages)
+
+        call_args = mock.call_args
+        system = call_args[1]["system"]
+        assert "Claude Code" in system
+
+
+def test_parse_response_text(provider):
+    """Should parse text response correctly."""
+    response = {
+        "content": [{"type": "text", "text": "Hello world"}],
+        "stop_reason": "end_turn",
+        "usage": {"input_tokens": 10, "output_tokens": 5},
+    }
+    result = provider._parse_response(response)
+    assert result.content == "Hello world"
+    assert result.finish_reason == "end_turn"
+    assert result.usage["prompt_tokens"] == 10
+
+
+def test_parse_response_tool_calls(provider):
+    """Should parse tool call response correctly."""
+    response = {
+        "content": [
+            {"type": "tool_use", "id": "call_1", "name": "read_file", "input": {"path": "/tmp/test"}}
+        ],
+        "stop_reason": "tool_use",
+        "usage": {"input_tokens": 10, "output_tokens": 5},
+    }
+    result = provider._parse_response(response)
+    assert len(result.tool_calls) == 1
+    assert result.tool_calls[0].name == "read_file"
+    assert result.tool_calls[0].arguments == {"path": "/tmp/test"}
+
+
+def test_convert_tools_to_anthropic(provider):
+    """Should convert OpenAI-format tools to Anthropic format."""
+    openai_tools = [
+        {
+            "type": "function",
+            "function": {
+                "name": "read_file",
+                "description": "Read a file",
+                "parameters": {"type": "object", "properties": {"path": {"type": "string"}}}
+            }
+        }
+    ]
+    anthropic_tools = provider._convert_tools_to_anthropic(openai_tools)
+    assert len(anthropic_tools) == 1
+    assert anthropic_tools[0]["name"] == "read_file"
+    assert "input_schema" in anthropic_tools[0]

tests/test_cli_oauth.py

@@ -0,0 +1,55 @@
+"""Test OAuth CLI commands."""
+import pytest
+import tempfile
+from pathlib import Path
+from typer.testing import CliRunner
+from nanobot.cli.oauth import oauth_app
+
+
+@pytest.fixture
+def runner():
+    return CliRunner()
+
+
+def test_oauth_login_help(runner):
+    """Login command should have help text."""
+    result = runner.invoke(oauth_app, ["login", "--help"])
+    assert result.exit_code == 0
+    assert "token" in result.output.lower()
+
+
+def test_oauth_status_no_credentials(runner, tmp_path, monkeypatch):
+    """Status should show no credentials when none exist."""
+    monkeypatch.setenv("HOME", str(tmp_path))
+    result = runner.invoke(oauth_app, ["status"])
+    assert result.exit_code == 0
+    assert "No OAuth credentials" in result.output
+
+
+def test_oauth_login_and_status(runner, tmp_path, monkeypatch):
+    """Login should save credentials, status should show them."""
+    monkeypatch.setenv("HOME", str(tmp_path))
+    result = runner.invoke(oauth_app, ["login", "--token", "sk-ant-oat01-test-xxx"])
+    assert result.exit_code == 0
+    assert "Successfully saved" in result.output
+
+    result = runner.invoke(oauth_app, ["status"])
+    assert result.exit_code == 0
+    assert "sk-ant-oat01-test-x" in result.output
+
+
+def test_oauth_logout(runner, tmp_path, monkeypatch):
+    """Logout should remove credentials."""
+    monkeypatch.setenv("HOME", str(tmp_path))
+    runner.invoke(oauth_app, ["login", "--token", "sk-ant-oat01-test-xxx"])
+    result = runner.invoke(oauth_app, ["logout"])
+    assert result.exit_code == 0
+    assert "Removed" in result.output
+
+
+def test_oauth_login_invalid_token(runner, tmp_path, monkeypatch):
+    """Login should reject non-OAuth tokens."""
+    monkeypatch.setenv("HOME", str(tmp_path))
+    result = runner.invoke(oauth_app, ["login", "--token", "sk-ant-api03-regular"])
+    assert result.exit_code == 0
+    assert "Invalid token" in result.output

tests/test_config_oauth_integration.py

@@ -0,0 +1,65 @@
+"""Test OAuth store integration with config loading."""
+import json
+import pytest
+import tempfile
+from pathlib import Path
+from nanobot.config.loader import load_config
+from nanobot.config.oauth_store import OAuthStore
+from nanobot.config.schema import OAuthCredentials
+
+
+def test_oauth_token_injected_into_config(tmp_path, monkeypatch):
+    """OAuth token from store should be injected into provider api_key."""
+    # Create a minimal config file (no api key set)
+    config_path = tmp_path / "config.json"
+    config_path.write_text(json.dumps({
+        "agents": {"defaults": {"model": "anthropic/claude-opus-4-5"}},
+        "providers": {"anthropic": {"apiKey": ""}}
+    }))
+
+    # Save OAuth credentials
+    store = OAuthStore(tmp_path)
+    creds = OAuthCredentials(access_token="sk-ant-oat01-test-inject")
+    store.save("anthropic", creds)
+
+    # Monkeypatch get_config_path to use our tmp dir
+    monkeypatch.setattr("nanobot.config.loader.get_config_path", lambda: config_path)
+    # Monkeypatch the OAuth store path
+    monkeypatch.setattr("nanobot.config.loader._get_oauth_store_dir", lambda: tmp_path)
+
+    config = load_config(config_path)
+
+    assert config.providers.anthropic.api_key == "sk-ant-oat01-test-inject"
+
+
+def test_config_without_oauth_unchanged(tmp_path, monkeypatch):
+    """Config without OAuth store should load normally."""
+    config_path = tmp_path / "config.json"
+    config_path.write_text(json.dumps({
+        "providers": {"anthropic": {"apiKey": "sk-ant-api03-regular"}}
+    }))
+
+    monkeypatch.setattr("nanobot.config.loader._get_oauth_store_dir", lambda: tmp_path / "nonexistent")
+
+    config = load_config(config_path)
+
+    assert config.providers.anthropic.api_key == "sk-ant-api03-regular"
+
+
+def test_oauth_does_not_overwrite_existing_key(tmp_path, monkeypatch):
+    """If user already has an API key, OAuth should still override (OAuth takes priority)."""
+    config_path = tmp_path / "config.json"
+    config_path.write_text(json.dumps({
+        "providers": {"anthropic": {"apiKey": "sk-ant-api03-existing"}}
+    }))
+
+    store = OAuthStore(tmp_path)
+    creds = OAuthCredentials(access_token="sk-ant-oat01-oauth-wins")
+    store.save("anthropic", creds)
+
+    monkeypatch.setattr("nanobot.config.loader._get_oauth_store_dir", lambda: tmp_path)
+
+    config = load_config(config_path)
+
+    # OAuth token takes priority over existing API key
+    assert config.providers.anthropic.api_key == "sk-ant-oat01-oauth-wins"

tests/test_oauth_config.py

@@ -0,0 +1,48 @@
+"""Test OAuth configuration schema."""
+import pytest
+from nanobot.config.schema import ProviderConfig, OAuthCredentials
+
+
+def test_provider_config_has_oauth_fields():
+    """ProviderConfig should have oauth_credentials field."""
+    config = ProviderConfig(api_key="test")
+    assert hasattr(config, "oauth_credentials")
+    assert config.oauth_credentials is None
+
+
+def test_oauth_credentials_model():
+    """OAuthCredentials should store token, refresh, expiry."""
+    creds = OAuthCredentials(
+        access_token="sk-ant-oat01-xxx",
+        refresh_token="rt_xxx",
+        expires_at=1234567890,
+        token_type="oauth"
+    )
+    assert creds.access_token.startswith("sk-ant-oat")
+    assert creds.is_oauth_token is True
+
+
+def test_oauth_credentials_expiry_check():
+    """OAuthCredentials should detect expired tokens."""
+    import time
+    expired = OAuthCredentials(
+        access_token="sk-ant-oat01-xxx",
+        expires_at=int(time.time()) - 3600  # 1 hour ago
+    )
+    assert expired.is_expired is True
+
+    valid = OAuthCredentials(
+        access_token="sk-ant-oat01-xxx",
+        expires_at=int(time.time()) + 3600  # 1 hour from now
+    )
+    assert valid.is_expired is False
+
+
+def test_oauth_credentials_no_expiry():
+    """Setup tokens with expires_at=0 should never be expired."""
+    creds = OAuthCredentials(
+        access_token="sk-ant-oat01-xxx",
+        expires_at=0  # No expiry (setup-token)
+    )
+    assert creds.is_expired is False
+    assert creds.expires_soon is False

tests/test_oauth_store.py

@@ -0,0 +1,55 @@
+"""Test OAuth credential storage."""
+import pytest
+import tempfile
+from pathlib import Path
+from nanobot.config.oauth_store import OAuthStore
+from nanobot.config.schema import OAuthCredentials
+
+
+@pytest.fixture
+def temp_store():
+    """Create store with temp directory."""
+    with tempfile.TemporaryDirectory() as tmpdir:
+        yield OAuthStore(Path(tmpdir) / ".nanobot")
+
+
+def test_save_and_load_credentials(temp_store):
+    """Should save and load OAuth credentials."""
+    creds = OAuthCredentials(
+        access_token="sk-ant-oat01-xxx",
+        refresh_token="rt_xxx",
+        expires_at=1234567890
+    )
+
+    temp_store.save("anthropic", creds)
+    loaded = temp_store.load("anthropic")
+
+    assert loaded is not None
+    assert loaded.access_token == creds.access_token
+    assert loaded.refresh_token == creds.refresh_token
+
+
+def test_load_nonexistent_returns_none(temp_store):
+    """Should return None for missing credentials."""
+    assert temp_store.load("nonexistent") is None
+
+
+def test_delete_credentials(temp_store):
+    """Should delete saved credentials."""
+    creds = OAuthCredentials(access_token="sk-ant-oat01-xxx")
+    temp_store.save("anthropic", creds)
+    assert temp_store.delete("anthropic") is True
+    assert temp_store.load("anthropic") is None
+
+
+def test_delete_nonexistent_returns_false(temp_store):
+    """Should return False when deleting missing credentials."""
+    assert temp_store.delete("nonexistent") is False
+
+
+def test_file_permissions(temp_store):
+    """Credentials file should have restricted permissions."""
+    creds = OAuthCredentials(access_token="sk-ant-oat01-xxx")
+    temp_store.save("anthropic", creds)
+    perms = oct(temp_store.file_path.stat().st_mode)[-3:]
+    assert perms == "600"

tests/test_oauth_utils.py

@@ -0,0 +1,28 @@
+"""Test OAuth utility functions."""
+import pytest
+from nanobot.providers.oauth_utils import is_oauth_token, get_auth_headers
+
+
+def test_is_oauth_token_detects_oat():
+    """Should detect sk-ant-oat tokens as OAuth."""
+    assert is_oauth_token("sk-ant-oat01-buSdhCH2XEkebW7ZQZTvGqH5EwAFh4u52LrdJhAP") is True
+    assert is_oauth_token("sk-ant-api03-regularkey") is False
+    assert is_oauth_token("") is False
+    assert is_oauth_token(None) is False
+
+
+def test_get_auth_headers_oauth():
+    """OAuth tokens should use Authorization: Bearer."""
+    headers = get_auth_headers("sk-ant-oat01-xxx", is_oauth=True)
+    assert "Authorization" in headers
+    assert headers["Authorization"] == "Bearer sk-ant-oat01-xxx"
+    assert "x-api-key" not in headers
+    assert headers["anthropic-beta"] == "claude-code-20250219,oauth-2025-04-20"
+
+
+def test_get_auth_headers_api_key():
+    """Regular API keys should use x-api-key."""
+    headers = get_auth_headers("sk-ant-api03-xxx", is_oauth=False)
+    assert "x-api-key" in headers
+    assert headers["x-api-key"] == "sk-ant-api03-xxx"
+    assert "Authorization" not in headers

tests/test_provider_factory.py

@@ -0,0 +1,32 @@
+"""Test provider factory with OAuth support."""
+import pytest
+from nanobot.providers import create_provider
+from nanobot.providers.anthropic_oauth import AnthropicOAuthProvider
+from nanobot.providers.litellm_provider import LiteLLMProvider
+
+
+def test_create_provider_oauth_token():
+    """OAuth tokens should create AnthropicOAuthProvider."""
+    provider = create_provider(
+        api_key="sk-ant-oat01-test-token",
+        model="anthropic/claude-opus-4-5"
+    )
+    assert isinstance(provider, AnthropicOAuthProvider)
+
+
+def test_create_provider_regular_key():
+    """Regular API keys should create LiteLLMProvider."""
+    provider = create_provider(
+        api_key="sk-ant-api03-regular-key",
+        model="anthropic/claude-opus-4-5"
+    )
+    assert isinstance(provider, LiteLLMProvider)
+
+
+def test_create_provider_openrouter():
+    """OpenRouter keys should create LiteLLMProvider."""
+    provider = create_provider(
+        api_key="sk-or-v1-xxx",
+        model="anthropic/claude-opus-4-5"
+    )
+    assert isinstance(provider, LiteLLMProvider)

tests/test_registry_oauth.py

@@ -0,0 +1,21 @@
+"""Test OAuth detection in provider registry."""
+import pytest
+from nanobot.providers.registry import should_use_oauth_provider
+
+
+def test_should_use_oauth_for_oat_token():
+    """OAuth provider should be used for sk-ant-oat tokens."""
+    assert should_use_oauth_provider("sk-ant-oat01-xxx", "anthropic/claude-opus-4-5") is True
+    assert should_use_oauth_provider("sk-ant-oat01-xxx", "claude-sonnet-4") is True
+
+
+def test_should_not_use_oauth_for_regular_key():
+    """Regular API keys should not use OAuth provider."""
+    assert should_use_oauth_provider("sk-ant-api03-xxx", "claude-opus-4-5") is False
+    assert should_use_oauth_provider("sk-or-v1-xxx", "anthropic/claude-opus-4-5") is False
+
+
+def test_should_not_use_oauth_for_non_anthropic():
+    """Non-Anthropic models should not use OAuth provider."""
+    assert should_use_oauth_provider("sk-ant-oat01-xxx", "gpt-4") is False
+    assert should_use_oauth_provider("sk-ant-oat01-xxx", "deepseek/deepseek-chat") is False