Control what your
AI agents can do.
OCC is a system where computations are only executable if they are authorized by a previously committed, cryptographically bound policy.
Available now
17 integrationsMCP (any server)
Default-deny policy enforcement for ANY MCP server
npx occ-mcp-proxy --wrap npx <any-mcp-server>Claude Desktop
Policy-enforced MCP servers in Claude Desktop
Add to claude_desktop_config.jsonCursor
Policy-controlled AI edits in your IDE
Add to .cursor/mcp.jsonOpenAI
Policy enforcement for OpenAI function calls
npm install occ-openaiOpenAI Agents
Policy-controlled agent tool execution
pip install occ-openai-agentsLangChain
Policy enforcement for LangChain tool calls
pip install occ-langchainLangGraph
Policy-controlled node execution in LangGraph
npm install occ-langgraphVercel AI SDK
Policy enforcement middleware for Vercel AI
npm install occ-vercelCrewAI
Policy-enforced tools for CrewAI agents
pip install occ-crewaiGoogle Gemini
Policy enforcement for Gemini function calling
pip install occ-geminiGoogle ADK
Policy-controlled Google ADK agents
pip install occ-google-adkLlamaIndex
Policy enforcement for LlamaIndex agents
pip install occ-llamaindexAutoGen
Policy-enforced multi-agent AutoGen workflows
pip install occ-autogenOpenClaw
Policy-controlled local AI with 20+ platforms
pip install occ-openclawMastra
Policy enforcement for Mastra AI agents
npm install occ-mastraCloudflare Workers
Policy enforcement at the edge on Cloudflare Workers
npm install occ-cloudflareGitHub Actions
Verify execution proofs in CI
uses: mikeargento/occ-verify-action@v1Live Proofs
View all in ExplorerEvery agent action that executes through OCC produces a proof. The proof authorized the action and is the verifiable record that it happened. Click to expand.
Orchestrators
Agent platformsOCC plugs into multi-agent orchestration platforms. Define a policy, and every tool call must be authorized by it. Revoke a permission, and the proof record updates accordingly. The proof that allowed the action and the proof that it happened are the same object.
Paperclip
Agent control plane with policy enforcement, revocation, and TEE-attested proofs on every action
Composio
250+ tool integrations with cryptographic control on every call
LangSmith
Controlled traces with cryptographic authorization for LangChain
Julep
Stateful agent workflows with cryptographic control at every step
AgentOps
Agent observability with cryptographic control and audit
Relevance AI
No-code agent builder with cryptographic control
Letta (MemGPT)
Long-term memory agents with cryptographic control
SuperAGI
Autonomous agent framework with OCC control layer
Don't see your framework?
OCC is open source. Add cryptographic control to any tool-calling framework in minutes. Contributions welcome.