GigAI

GigAI is a local, user-controlled agent runtime. Configuration, credentials and work state stay under a home directory you choose, and every model call, review and result is inspectable rather than taken on trust. Work is packaged as Gigs: goal graphs with sealed inputs, approval gates and a run journal. It runs on the Codex or Claude CLI you already use, is Apache-2.0, and installs with uv tool install gigai.

The guardrails live in code, not in the prompt. Contact details are stripped from every payload before it reaches a model, and tests capture those payloads to prove it. Resume lines are picked from your master word for word, each with its source. Assessments are measured against frozen, gold-keyed cases with a blind judge, and the results drive the prompt and code fixes.

One Gig is published so far: Scout. It runs as a local service with a browser UI and an API, so you can drive it yourself or hand it to your own agent. In the background it checks about 15,000 public job boards eight times a day.

Built with

Python React Typst Codex CLI Claude CLI Apache-2.0