Policy as code for AI agents: your rules are files in git, enforced at the tool call across 43 agent harnesses.
Website • Documentation • Main Repository • Slack
Observability tools log what an agent did after it leaked a secret, deleted files or looped on a paid API.
Intutic is an active circuit breaker at the tool call. Your rules are Markdown files in .intutic/sops/. A gate installed in each agent harness checks every tool call against them before it runs, and a local proxy checks the model traffic. Dangerous commands are refused, secrets are redacted, and loops are steered or stopped. The policy check is a rule decision, not a model call; rules that need a judgement can ask your own self-hosted model.
| Repository | Description | Stars / Status |
|---|---|---|
🛡️ intutic/intutic |
The open-core repository: the Rust proxy, @intutic/cli, the tool-call gates (@intutic/gate for TypeScript, intutic-clawde for Python), the @intutic/clawde client SDK, the MCP governance proxy and the WASM rules SDK. |
flowchart TD
subgraph Dev[" 💻 Developer machine "]
Agent["🤖 AI coding agent<br><i>(Claude Code, Cursor, Codex, Aider, LangGraph, …)</i>"]
SOP["📝 SOP rules<br><i>(.intutic/sops/)</i>"]
Gate["🚦 Tool-call gate<br><i>(runs before each tool call)</i>"]
end
subgraph Proxy[" ⚡ Intutic proxy (:4000) "]
Checks["🔒 Checks on each request and response<br><i>SOP rules · WASM rules · DLP · anomaly detectors · budget</i>"]
Verdicts["Verdicts<br><code>BYPASS</code> | <code>HIJACK</code> | <code>REASK</code> | <code>KILL</code>"]
end
Providers["🌐 LLM providers<br><i>Anthropic · OpenAI · Gemini · Bedrock · Vertex · Azure · …</i>"]
Agent -->|1. Model request| Proxy
Proxy -->|2. Allowed request| Providers
Providers -->|3. Response with tool calls| Proxy
Proxy -->|4. Response, or refusal| Agent
Agent -->|5. Each tool call| Gate
SOP -->|Compiled into the gate and reloaded by the proxy| Checks
SOP --> Gate
npm install -g @intutic/cli @intutic/proxyintutic init
intutic gates installThis detects the coding agents in your project and installs a blocking gate in each.
intutic startNo account, no Docker, no config file. Everything runs on your machine.
# Anthropic SDK and Claude Code append /v1/messages themselves — host only.
export ANTHROPIC_BASE_URL="http://localhost:4000"
export OPENAI_BASE_URL="http://localhost:4000/v1"Intutic ships 43 harness adapters:
- Native hook gates (no change to your agent's code): Claude Code, Cursor, Windsurf, Cline, Codex, Goose, GitHub Copilot, OpenCode and more.
- SDK-gated frameworks (an in-process gate you add to your agent): LangGraph, LangChain, CrewAI, AutoGen, AG2, Google ADK, OpenAI Agents SDK, Pydantic AI, smolagents, AWS Strands Agents, Microsoft Agent Framework, Mastra, Vercel AI SDK and more.
- Through the proxy alone: Aider, Claude Desktop, Roo Code, Continue (these also use the MCP governance proxy where they have MCP), and anything else that honours
OPENAI_BASE_URLorANTHROPIC_BASE_URL.
The full list is in the main repository.
- 📖 Documentation Portal: docs.intutic.ai
- 🌐 Official Website: intutic.ai
- 💬 Slack Community: Join Intutic Slack
- 🐤 Follow on X: @IntuticAI
- 💼 LinkedIn Page: Intutic on LinkedIn
- 📧 Contact Support: support@intutic.ai
© 2026 Intutic Community. All rights reserved. Licensed under the MIT License.