agenthandler 0.92.0__tar.gz → 0.93.0__tar.gz

{agenthandler-0.92.0 → agenthandler-0.93.0}/.gitignore

@@ -27,11 +27,6 @@ htmlcov/
 node_modules/
 openclaw-plugin/dist/
 
-# Stray installers / large binaries
-npm/*.sh
-*.pkg
-*.dmg
-
 # Environment
 .env
 .env.*

{agenthandler-0.92.0 → agenthandler-0.93.0}/CHANGELOG.md

@@ -12,9 +12,14 @@ All notable changes to AgentHandler are documented here.
 - Test count badge in README.
 
 ### Changed
-- **mypy --strict** enabled across all 29 source files with zero errors. Full type annotations on every function.
+- **Rebrand**: Clawboss → AgentHandler. `pip install agenthandler`.
+- **MCP server mode**: `SupervisedMCPServer` exposes tools via Model Context Protocol with full supervision. `pip install agenthandler[mcp]`.
+- **mypy --strict** enabled across all 30 source files with zero errors. Full type annotations on every function.
 - Python 3.13 compatibility: replaced deprecated `asyncio.get_event_loop()` in approval tests.
 
+### Removed
+- `openclaw-plugin/` directory — superseded by native MCP server mode. The Python-side `OpenClawBridge` remains for backward compatibility.
+
 ## [0.91.0] — 2026-06-10
 
 ### Added

{agenthandler-0.92.0 → agenthandler-0.93.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: agenthandler
-Version: 0.92.0
+Version: 0.93.0
 Summary: A way of controlling deployed agents to keep your (and their) sanity.
 Project-URL: Homepage, https://github.com/arunvenkatadri/AgentHandler
 Project-URL: Repository, https://github.com/arunvenkatadri/AgentHandler
@@ -29,6 +29,8 @@ Provides-Extra: kafka
 Requires-Dist: aiokafka>=0.10; extra == 'kafka'
 Provides-Extra: kinesis
 Requires-Dist: aioboto3>=12.0; extra == 'kinesis'
+Provides-Extra: mcp
+Requires-Dist: mcp>=1.0; extra == 'mcp'
 Provides-Extra: redis
 Requires-Dist: redis>=5.0; extra == 'redis'
 Provides-Extra: server
@@ -56,9 +58,11 @@ Description-Content-Type: text/markdown
 [![Zero Dependencies](https://img.shields.io/badge/dependencies-0-brightgreen.svg)]()
 [![Tests](https://img.shields.io/badge/tests-658-brightgreen.svg)]()
 
-**Stop your AI agents from going rogue and set your long acting agents up for success.** AgentHandler wraps tool calls with timeouts, budgets, circuit breakers, and audit logging so one bad tool call doesn't drain your wallet or loop forever.
+*Every agent needs a handler.*
 
-Zero dependencies. Works with **any agent framework** — LangChain, CrewAI, AutoGen, OpenClaw, your own custom loop, whatever. Just wrap your tool calls. Includes durable sessions that survive restarts, a REST control plane, and a [dashboard](#dashboard) for managing everything in one place.
+**AgentHandler is the governance layer for AI agents.** Wrap tool calls with timeouts, budgets, circuit breakers, guardrails, and audit logging — so one bad tool call doesn't drain your wallet or loop forever. MCP-native, with adapters for OpenAI Agents SDK and Claude Agent SDK.
+
+Zero dependencies. Works with **any agent framework** — OpenAI Agents SDK, Claude Agent SDK, Mastra, LangChain, CrewAI, your own custom loop, whatever. Just wrap your tool calls. Includes durable sessions that survive restarts, a REST control plane, MCP server mode, and a [dashboard](#dashboard) for managing everything in one place.
 
 > **AgentHandler is built for long-duration agents** — agents that run for hours or days, not minutes. Read [the manifesto](docs/manifesto.md) for the full vision.
 
@@ -178,11 +182,12 @@ The Sessions tab connects to the REST control plane (`uvicorn agenthandler.serve
 AgentHandler doesn't care what framework you use. It supervises tool calls — any async or sync callable. If your agent calls tools, AgentHandler can wrap them.
 
 ```python
-# LangChain? Wrap your tools.
-# CrewAI? Wrap your tools.
-# AutoGen? Wrap your tools.
+# OpenAI Agents SDK? Wrap your tools.
+# Claude Agent SDK? Wrap your tools.
+# Mastra? Wrap your tools.
+# LangChain / CrewAI? Wrap your tools.
+# MCP server? Built-in (see below).
 # Custom loop? Wrap your tools.
-# OpenClaw? There's a built-in bridge (see below).
 
 result = await supervisor.call("my_tool", my_tool_fn, **kwargs)
 ```
@@ -792,46 +797,42 @@ AgentHandler is designed to supervise untrusted agent behavior. The stateful ses
 
 **Sessions can expire.** Call `SqliteStore.delete_expired(max_age_seconds)` to clean up old sessions.
 
-## OpenClaw integration
+## MCP server mode
+
+Expose supervised tools via Model Context Protocol. Any MCP client — Claude Desktop, Claude Code, Cursor — gets governed tool access with full supervision.
 
-AgentHandler includes a built-in bridge for [OpenClaw](https://github.com/openclaw/openclaw). Expose your supervised tools to OpenClaw over HTTP — all supervision (timeouts, budgets, circuit breakers) applies automatically.
+```bash
+pip install agenthandler[mcp]
+```
 
 ```python
-from agenthandler import OpenClawBridge, Skill, ToolDefinition, ToolParameter
-
-# Define your skill with tools and supervision limits
-skill = Skill(
-    name="web_research",
-    description="Research topics on the web",
-    tools=[
-        ToolDefinition(
-            name="web_search",
-            description="Search the web",
-            parameters=[
-                ToolParameter(name="query", type="string",
-                              description="Search query", required=True),
-            ],
-        ),
+from agenthandler import SupervisedMCPServer, UrlGuard, RecursionDetector
+
+server = SupervisedMCPServer(
+    name="my-tools",
+    tools={
+        "search": search_fn,
+        "write": write_fn,
+        "send_email": email_fn,
+    },
+    policy={
+        "tool_timeout": 30,
+        "token_budget": 50000,
+        "require_confirm": ["send_email"],
+    },
+    pre_guardrails=[
+        UrlGuard(allowlist=["*.example.com"]),
+        RecursionDetector(max_repeats=3),
     ],
-    supervision={"tool_timeout": 15, "max_iterations": 5, "token_budget": 10000},
 )
 
-# Start the bridge
-bridge = OpenClawBridge(port=9229)
-bridge.register_skill(skill, {"web_search": my_search_fn})
-bridge.serve()  # GET /tools, POST /execute/{name}
+server.run()  # stdio for Claude Desktop
+# server.run(transport="streamable-http")  # network mode
 ```
 
-Then install the TypeScript plugin from `openclaw-plugin/` into OpenClaw. The plugin auto-discovers tools from the bridge and registers them. See `examples/openclaw_bridge.py` for a full working example.
-
-You can also convert schemas without running a bridge:
+Every tool call through MCP goes through the Supervisor — timeouts, budgets, circuit breakers, PII redaction, guardrails, approvals, audit, cost tracking. A capability manifest is available at `agenthandler://manifest`.
 
-```python
-from agenthandler import to_openclaw_tool_schema, to_openclaw_manifest
-
-schema = to_openclaw_tool_schema(tool_def)    # OpenClaw JSON Schema format
-manifest = to_openclaw_manifest(skill)         # openclaw.plugin.json content
-```
+See `examples/mcp_server.py` for a full working example.
 
 ## Policy from config
 

{agenthandler-0.92.0 → agenthandler-0.93.0}/README.md

@@ -13,9 +13,11 @@
 [![Zero Dependencies](https://img.shields.io/badge/dependencies-0-brightgreen.svg)]()
 [![Tests](https://img.shields.io/badge/tests-658-brightgreen.svg)]()
 
-**Stop your AI agents from going rogue and set your long acting agents up for success.** AgentHandler wraps tool calls with timeouts, budgets, circuit breakers, and audit logging so one bad tool call doesn't drain your wallet or loop forever.
+*Every agent needs a handler.*
 
-Zero dependencies. Works with **any agent framework** — LangChain, CrewAI, AutoGen, OpenClaw, your own custom loop, whatever. Just wrap your tool calls. Includes durable sessions that survive restarts, a REST control plane, and a [dashboard](#dashboard) for managing everything in one place.
+**AgentHandler is the governance layer for AI agents.** Wrap tool calls with timeouts, budgets, circuit breakers, guardrails, and audit logging — so one bad tool call doesn't drain your wallet or loop forever. MCP-native, with adapters for OpenAI Agents SDK and Claude Agent SDK.
+
+Zero dependencies. Works with **any agent framework** — OpenAI Agents SDK, Claude Agent SDK, Mastra, LangChain, CrewAI, your own custom loop, whatever. Just wrap your tool calls. Includes durable sessions that survive restarts, a REST control plane, MCP server mode, and a [dashboard](#dashboard) for managing everything in one place.
 
 > **AgentHandler is built for long-duration agents** — agents that run for hours or days, not minutes. Read [the manifesto](docs/manifesto.md) for the full vision.
 
@@ -135,11 +137,12 @@ The Sessions tab connects to the REST control plane (`uvicorn agenthandler.serve
 AgentHandler doesn't care what framework you use. It supervises tool calls — any async or sync callable. If your agent calls tools, AgentHandler can wrap them.
 
 ```python
-# LangChain? Wrap your tools.
-# CrewAI? Wrap your tools.
-# AutoGen? Wrap your tools.
+# OpenAI Agents SDK? Wrap your tools.
+# Claude Agent SDK? Wrap your tools.
+# Mastra? Wrap your tools.
+# LangChain / CrewAI? Wrap your tools.
+# MCP server? Built-in (see below).
 # Custom loop? Wrap your tools.
-# OpenClaw? There's a built-in bridge (see below).
 
 result = await supervisor.call("my_tool", my_tool_fn, **kwargs)
 ```
@@ -749,46 +752,42 @@ AgentHandler is designed to supervise untrusted agent behavior. The stateful ses
 
 **Sessions can expire.** Call `SqliteStore.delete_expired(max_age_seconds)` to clean up old sessions.
 
-## OpenClaw integration
+## MCP server mode
+
+Expose supervised tools via Model Context Protocol. Any MCP client — Claude Desktop, Claude Code, Cursor — gets governed tool access with full supervision.
 
-AgentHandler includes a built-in bridge for [OpenClaw](https://github.com/openclaw/openclaw). Expose your supervised tools to OpenClaw over HTTP — all supervision (timeouts, budgets, circuit breakers) applies automatically.
+```bash
+pip install agenthandler[mcp]
+```
 
 ```python
-from agenthandler import OpenClawBridge, Skill, ToolDefinition, ToolParameter
-
-# Define your skill with tools and supervision limits
-skill = Skill(
-    name="web_research",
-    description="Research topics on the web",
-    tools=[
-        ToolDefinition(
-            name="web_search",
-            description="Search the web",
-            parameters=[
-                ToolParameter(name="query", type="string",
-                              description="Search query", required=True),
-            ],
-        ),
+from agenthandler import SupervisedMCPServer, UrlGuard, RecursionDetector
+
+server = SupervisedMCPServer(
+    name="my-tools",
+    tools={
+        "search": search_fn,
+        "write": write_fn,
+        "send_email": email_fn,
+    },
+    policy={
+        "tool_timeout": 30,
+        "token_budget": 50000,
+        "require_confirm": ["send_email"],
+    },
+    pre_guardrails=[
+        UrlGuard(allowlist=["*.example.com"]),
+        RecursionDetector(max_repeats=3),
     ],
-    supervision={"tool_timeout": 15, "max_iterations": 5, "token_budget": 10000},
 )
 
-# Start the bridge
-bridge = OpenClawBridge(port=9229)
-bridge.register_skill(skill, {"web_search": my_search_fn})
-bridge.serve()  # GET /tools, POST /execute/{name}
+server.run()  # stdio for Claude Desktop
+# server.run(transport="streamable-http")  # network mode
 ```
 
-Then install the TypeScript plugin from `openclaw-plugin/` into OpenClaw. The plugin auto-discovers tools from the bridge and registers them. See `examples/openclaw_bridge.py` for a full working example.
-
-You can also convert schemas without running a bridge:
+Every tool call through MCP goes through the Supervisor — timeouts, budgets, circuit breakers, PII redaction, guardrails, approvals, audit, cost tracking. A capability manifest is available at `agenthandler://manifest`.
 
-```python
-from agenthandler import to_openclaw_tool_schema, to_openclaw_manifest
-
-schema = to_openclaw_tool_schema(tool_def)    # OpenClaw JSON Schema format
-manifest = to_openclaw_manifest(skill)         # openclaw.plugin.json content
-```
+See `examples/mcp_server.py` for a full working example.
 
 ## Policy from config
 

{agenthandler-0.92.0 → agenthandler-0.93.0}/agenthandler/__init__.py

@@ -17,7 +17,7 @@ Usage:
         print(result.error)
 """
 
-__version__ = "0.92.0"
+__version__ = "0.93.0"
 
 from .approval import ApprovalQueue, ApprovalRequest, ApprovalStatus
 from .audit import AuditEntry, AuditLog, AuditOutcome, AuditPhase, AuditSink, JsonlAuditSink
@@ -54,6 +54,13 @@ from .policy import Action, OnFailure, Policy, ScopeRule, ToolScope
 from .redact import Redactor
 from .reflection import ReflectionCycle, ReflectionLoop, ReflectionResult
 from .replay import ReplayFrame, ReplaySummary, SessionReplay
+from .sdk_adapters import (
+    AgentHandlerMiddleware,
+    openai_guardrail_adapter,
+    supervised_tool_registry,
+    wrap_claude_tool,
+    wrap_openai_tool,
+)
 from .session import SessionManager
 from .skill import Skill, SkillStore, ToolDefinition, ToolParameter
 from .store import (
@@ -73,6 +80,16 @@ from .streams import (
 from .supervisor import SupervisedResult, Supervisor
 from .triggers import DbWatchEntry, Scheduler, WebhookTrigger
 
+try:
+    from .mcp_server import SupervisedMCPServer
+except ImportError:
+    pass
+
+try:
+    from .a2a import A2AAgentCard, A2AClient, A2ASkill, A2ASupervisedEndpoint, A2ATask, TaskState
+except ImportError:
+    pass
+
 __all__ = [
     "Policy",
     "OnFailure",
@@ -160,4 +177,19 @@ __all__ = [
     "SessionReplay",
     "ReplayFrame",
     "ReplaySummary",
+    # MCP
+    "SupervisedMCPServer",
+    # A2A
+    "A2AAgentCard",
+    "A2AClient",
+    "A2ASkill",
+    "A2ASupervisedEndpoint",
+    "A2ATask",
+    "TaskState",
+    # SDK adapters
+    "wrap_openai_tool",
+    "wrap_claude_tool",
+    "openai_guardrail_adapter",
+    "supervised_tool_registry",
+    "AgentHandlerMiddleware",
 ]