Appearance

IntuticThe circuit breaker for AI agents

Intutic sits in the tool-call path between your AI agents and production. Every file write, API call, and shell command passes through a <5ms evaluation chain — blocking bad actions and steering agentic loops in real time.

Get Started
View on GitHub
🔴

Circuit Breaker

Four enforcement actions — BYPASS, ENHANCE, HIJACK, KILL — evaluated per tool call in under 50ms. SOPs define the rules; the circuit breaker intercepts, rewrites, or blocks actions before they reach your infrastructure.

🔍

DLP & Threat Detection

Detect secrets in prompts, data exfiltration, prompt injection, tool abuse, token waste, SQL injection, and unauthorized operations. Threats are flagged and blocked before they leave the proxy.

💰

FinOps Ledger

Track every token, every model, every dollar. Per-model cost breakdowns, local session token metering, and customizable spending caps.

🔌

18 Harness Integrations

Works with Claude Code, Cursor, Windsurf, Aider, Antigravity, Codex, OpenHands, n8n, Cline, Roo Code, Continue, Claude Desktop, Goose, Open WebUI, OpenClaw, Hermes, Pi, and GitHub Copilot. Auto-detects your tooling and syncs governance rules to every agent.

Quick Install ​

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

# Start the local proxy gateway (no account needed)
intutic start

Your workspace is now governed. Every AI agent request flows through the Intutic proxy, where SOPs are enforced and traces are recorded.

What Gets Blocked ​

Intutic's circuit breaker evaluates every tool call against your policy stack. Here's what gets caught:

ThreatHow It's Blocked
🗂️ File system destructionrm -rf, mass deletes, and recursive overwrites are intercepted by path-pattern SOPs
🔐 Secrets in promptsAPI keys, tokens, and credentials are redacted before they reach the model
🏭 Production writesDatabase mutations against prod connections are blocked or routed to review queue
🔀 Unauthorized git opsForce pushes, branch deletions, and pushes to protected branches are killed
💸 Budget overrunsToken spend exceeding session ceilings triggers automatic session suspension
🌐 Unapproved API callsOutbound HTTP to non-allowlisted domains is blocked at the proxy layer
🧩 MCP tool violationsCalls to unapproved MCP servers or tools are intercepted before execution
💉 SQL injectionDestructive SQL patterns (DROP, TRUNCATE, DELETE without WHERE) are caught and blocked

Every blocked action generates an audit log entry with full context — who, what, when, and why it was stopped.

🐱 Agentic Custom Skill (Kitkat) ​

If you are using an agentic coding assistant (such as Claude Code, Cursor, Google Antigravity, or OpenAI Codex), you can download our pre-configured Kitkat Governance Skill to load rules directly into your workspace:

The circuit breaker for AI agents

© 2026 Intutic Community. All rights reserved.