AI-TC Open source plugin · Agent Harness Security

The security layer for
your agent harness

AKA AI-TC intercepts prompts and tool calls before they reach the model, scans them against your rule packs, and enforces policy — entirely on your machine, with zero backend.

Open source Local Auditable
aka session monitor local-first
  • readsrc/api/payments.tsallow
  • detectsecrets/aws-access-keyblock
  • editsrc/services/notify.tsallow
  • detectcore-pii/emailredact
  • detectcore-pii/ssnwarn
  • promptsupport-script.mdallow
store ~/.aka/data/aka.db local
Traffic flow
Prompts Tool calls Responses File reads AKA policy engine Monitor Warn Redact Block Exception
AI Traffic Control ai-tc

AKA AI-TC (AI Traffic Control) is an open-source control plane that sits between your data and AI tools — intercepting prompts, scanning for sensitive data, enforcing policy, and giving full control into what's being shared with AI models.

Open Source Local-first Claude Code Plugin

Get started in minutes

Install the plugin

The plugin + aka CLI write a local SQLite store (~/.aka/data/aka.db) and the OSS web-ui reads it directly — no backend, no Postgres, no account.

See the full Quickstart guide for the step-by-step walkthrough — cloning, installing the Claude Code plugin, triggering a detection, and opening the dashboard.

Terminal
# clone and set up
$ git clone https://github.com/akasecurity/ai-tc
$ cd ai-tc
$ pnpm setup
$ pnpm --filter @akasecurity/cli build

# launch the local dashboard
$ pnpm --filter @akasecurity/cli exec aka dashboard
Local architecture
your machine · no backend Claude Code hooks AKA plugin writes ~/.aka/data/aka.db reads Dashboard / CLI

Runs entirely on your machine

No backend, no cloud, no account

The plugin and aka CLI write straight to a local SQLite store at ~/.aka/data/aka.db, and the OSS dashboard reads that same file directly — no server to stand up, no Postgres, no account to create.

See the CLI guide for how the plugin, CLI, and dashboard all share that single local store.

7 rule packs · 100+ detections

Detect everything, enforce what matters

Every rule ships active under the log-only Monitor policy. Promote any detection to Warn, Redact, or Block from the dashboard — per rule, per category, or org-wide.

See the full list of built-in rules, or learn how to write your own.

Detection catalog
Monitor Warn Redact Block
  • Secrets21 rules
  • Secrets Infra13 rules
  • Core PII14 rules
  • Core Financial8 rules
  • Core PHI8 rules
  • Core Code Context8 rules
  • Code Flaws29 rules