Guardrails for the agent harness.

AI Traffic Control keeps regulated data on your machine and checks every prompt and tool call before it runs.

ai-tc enforced locally
prompt pci/card-number in pasted log redacted
tool call shell: rm -rf in /prod blocked
file read phi/patient-record in fixtures redacted
tool call fetch → untrusted model API blocked
response secret/aws-key in output warned
file read src/api/webhooks.ts allowed
PCI, PHI and PII never reach a model or a third party.

One control point for every event.

Prompts and tool inputs are checked before they reach the model. Responses and file reads are checked after. Each match becomes a finding, and policy decides the outcome.

prompttool callresponsefile read
ai-tc
rule packs · policy
monitorwarnredactblockexception

Runs next to the agent.

No backend to stand up and nothing leaves the machine to be scanned. Claude Code and Claude Desktop are supported today.

$ /plugin marketplace add akasecurity/marketplace
$ /plugin install ai-tc@akasecurity
$ /aka:setup

Build Your Own Security

or