Intutic — Policy as Code for Continuous Compliance & Continuous Enforcement for AI Agents

The circuit breaker for AI agents: your policies are files in git, enforced synchronously and in-process on every tool call across 42 agent harnesses.

Quickstart • Architecture • Key Features • Supported Harnesses • Docs

Existing AI observability tools (like LangSmith or Portkey) are passive. They record execution logs after an agent leaks a secret, deletes files, or loops into hundreds of dollars of API spend.

Intutic is an active circuit breaker. It sits in the tool-call path between your AI agents and local shell/production APIs. Every tool execution passes through an in-process policy evaluation chain — no model call, no network hop — blocking dangerous commands before they run and steering agentic loops in real time.

Intutic runs as a high-performance local or self-hosted proxy (written in Rust) alongside a lightweight bidirectional config sync daemon (sync-daemon):

flowchart TD
    subgraph DevEnvironment[" 💻 Developer Environment "]
        Agent["🤖 AI Coding Agent<br><i>(Claude Code, Cursor, Aider, LangGraph)</i>"]
        SOP["📝 Local SOP Rules<br><i>(CLAUDE.md / .cursorrules / SKILL.md)</i>"]
    end

    subgraph HotPathProxy[" ⚡ Intutic Hot-Path Proxy (:4000) "]
        Engine["🔒 WASM Policy Engine<br><i>(In-Process Evaluation)</i>"]
        DLP["🔐 Secret DLP & Masking"]
        PCAS["🛡️ PCAS Action Primitives<br><code>BYPASS</code> | <code>ENHANCE</code> | <code>HIJACK</code> | <code>REASK</code> | <code>KILL</code>"]
    end

    subgraph SyncDaemon[" 🔄 Sync Daemon "]
        Reconcile["Bidirectional Config Reconciler<br><i>(Harness Config Sync)</i>"]
    end

    subgraph UpstreamProviders[" 🌐 Upstream Providers "]
        Providers["Anthropic API / OpenAI / LiteLLM / Ollama"]
    end

    Agent -->|1. Tool Call / Prompt| HotPathProxy
    SOP -->|2. Rule Sync| Reconcile
    Reconcile -->|3. Hot-Reload Rules| Engine
    HotPathProxy -->|4. Clean Request| UpstreamProviders
    Engine -->|5. Block/Hijack Verdict| Agent

# Install global CLI and native Rust proxy binary
npm install -g @intutic/cli @intutic/proxy

# Or run the native proxy directly on-demand
npx @intutic/proxy

Run intutic connect inside your project folder. This boots the local high-speed Rust proxy on port 4000 and auto-detects installed coding assistants:

intutic connect

Point your favorite LLM client or agent framework to the local proxy:

export ANTHROPIC_BASE_URL="http://localhost:4000/v1"
export OPENAI_BASE_URL="http://localhost:4000/v1"

That's it! Your agent is now governed by real-time safety guardrails.

Every tool call and prompt evaluated by Intutic produces one of five PCAS Action Primitives:

 ┌──────────┐  ┌───────────┐  ┌────────────┐  ┌──────────┐  ┌──────────┐
 │  BYPASS  │  │  ENHANCE  │  │   HIJACK   │  │  REASK   │  │   KILL   │
 └────┬─────┘  └─────┬─────┘  └─────┬──────┘  └────┬─────┘  └────┬─────┘
      │              │              │              │              │
      ▼              ▼              ▼              ▼              ▼
 Direct Pass    Inject Safety  Redact Secrets    Refuse &      Hard-Abort
 (In-Process)    Context SOP    or Swap Args    Bounded Retry  Runaway Loop

1. BYPASS: Standard safe execution passes through natively (in-process, no added network hop).
2. ENHANCE: Inject contextual SOP prompt rules or architectural guidelines.
3. HIJACK: Substitute dangerous tool parameters or redact secrets on the fly.
4. REASK: Refuse the attempt and hand the reason back to the agent, which may retry a bounded number of times before the finding escalates to a block.
5. KILL: Hard-abort execution thread if an agent attempts destructive file/git ops or hits loop caps.

Intutic ships 42 harness adapters that are auto-detected and config-synced without modifying your agent's source code, and every one of them is a supported harness — no adapter carries an open support gap today (AutoGen's closed with its workbench gate, the Agentic Orchestrator's with OpenCode's own adapter, both in September 2026). Anything else that speaks an OpenAI- or Anthropic-compatible API is governed the same way by pointing its base URL at the proxy:

Intutic governance rules are written in standard Markdown files inside your repository root (CLAUDE.md, .cursorrules, or .windsurfrules). Intutic automatically syncs and enforces them in real time:

# Standard Operating Procedure (SOP): Safety Guardrails

## Rules
1. **No Secret Leaks**: Agents must never output raw API keys or passwords.
2. **File Boundaries**: Restrict file modifications to the current project directory.
3. **No Force Push**: Block `git push --force` on all branches.

## Denied Commands
- `rm -rf`
- `DROP TABLE`
- `TRUNCATE`

Because Intutic evaluates prompts pre-flight, you can run interactive governance commands directly inside your agent chat:

/intutic status   # View active session spend and compliance score
/intutic rules    # List active WASM & Markdown SOP rules

• 📖 Developer Portal & API Reference: docs.intutic.ai
• 🏗️ Architecture & Contributor Guide: info.md
• 🛠️ WASM Rules SDK: packages/wasm-sdk/
• 📦 NPM Package Suite:
  • @intutic/cli — Developer onboarding CLI
  • @intutic/proxy — Native Rust proxy binary wrapper
  • @intutic/clawde — Programmatic TypeScript client SDK
  • @intutic/mcp-governance-proxy — Stdio JSON-RPC MCP proxy
  • @intutic/shared-types — Zod-validated shared TypeScript types

If you find Intutic useful, please give us a star on GitHub! It helps us support more agent harnesses and policy engines.

For custom VPC deployments, enterprise-grade SSO/SAML, dedicated SLA support, or team compliance auditing, visit intutic.ai or contact us at support@intutic.ai.

This project is licensed under the MIT License.