Can you tell me what they did yesterday? Exactly?
Right now, your AI agents are spending your money, accessing your data, and making decisions on your behalf. You have no spending limits, no audit trail, and no rules about what they can or can't do. When something goes wrong — and it will — you won't know until the damage is done.
If you're running multiple AI agents, one of these has already happened to you — or it will.
An AI agent burned through $800 in API calls in a single afternoon. Nobody noticed until the invoice arrived at the end of the month.
An agent replied to a customer with completely hallucinated information. The client saw it before any human reviewed it. Trust: gone.
Your compliance officer asks: "Who authorized this AI to access customer data?" You open a blank spreadsheet. You have no answer.
SIDJUA is a governance layer that sits between you and your AI agents. Every action gets checked before it runs.
Per-agent budget caps, enforced automatically. Set a limit, forget about it. No more surprise API bills.
Every action is checked against your rules before it executes. Violations are blocked, not just logged.
Know exactly what every agent did, when, and why. Investigate any incident in 30 seconds flat.
SIDJUA intercepts every agent action and runs it through your rules before it ever executes.
Any AI agent — Claude, GPT, Gemini, local models — wants to do something.
5-step governance pipeline validates against your policies and budget limits.
Compliant actions proceed. Violations are stopped before they cause damage.
Full audit trail. Every action, every decision, every cost — searchable and exportable.
git clone https://github.com/GoetzKohlberg/sidjua.git cd sidjua docker compose up -d docker exec -it sidjua sidjua init docker exec -it sidjua sidjua chat guide
Requires Docker and Docker Compose. See full docs on GitHub.
2 minutes. No AI voiceover. Just the problem, the story, and the fix.
"The architecture heavily isolates governance from execution. SQLite is an excellent choice for the air-gapped, self-hosted product promise."
— Gemini 3.1 Pro, Architecture Audit Round 2 (March 2026)
Triple-provider audit system: Gemini + DeepSeek + independent review. All findings resolved.
Three commands. Four gigs of RAM. Complete AI agent governance.
Skip the "cool project bro" — give us structured feedback that actually helps. It takes 2 minutes and works like a focused survey, not a comment thread.
Give Feedback