akasecurity/ai-tc

17 stars · Last commit 2026-11-09

Guardrails for the agent harness. Regulated data stays on your machine, and every prompt and tool call is scanned before it runs.

README preview

<picture>
  <source media="(prefers-color-scheme: dark)" srcset="assets/cover-dark.svg">
  <img src="assets/cover-light.svg" alt="AI Traffic Control (ai-tc) by AKA Security. Illustration, shown on a demo session: a traffic-control plaza seen from above. Prompts, tool calls, responses and file reads arrive as vehicles, pass under the ai-tc policy engine gate, and leave in five lanes: monitor, warn, redact, block (barrier down, lane closed) and exception (set apart)." width="100%">
</picture>

<p align="center">
  <a href="https://akasecurity.github.io/ai-tc-docs/"><b>Docs</b></a> &nbsp;·&nbsp;
  <a href="#install"><b>Install</b></a> &nbsp;·&nbsp;
  <a href="https://akasecurity.github.io/ai-tc-docs/rules/built-in/"><b>Rules</b></a> &nbsp;·&nbsp;
  <a href="SECURITY.md"><b>Security</b></a>
</p>

<h1 align="center">AI Traffic Control</h1>

<p align="center"><b>Every event in an agent session runs through one control point before it takes effect.</b></p>

<p align="center">Prompts, tool calls, responses and file reads, scanned against your rule packs on your machine.<br>Open source, by <a href="https://akasecurity.io">AKA Security</a>. We secure agent harnesses at the source.</p>

<p align="center">
  <img alt="Open source" src="https://img.shields.io/badge/Open_source-232F3E?style=flat-square">

View full repository on GitHub →