Governed autonomy for serious repositoriesGoverned autonomy
arbiterForge builds the governance layer for AI coding agents: hard gates instead of hopeful prompts, durable project context, and an audit trail that survives whichever host you use.
You decide. The gates enforce.
Evidence
This is not a description of enforcement. It is a replay of it. Both transcripts below come from direct invocations of the shipped pre-bash.py hook. Same file, same digest, real exit codes.
$ git push origin main
BLOCKED [H-01]: Pushing to a protected branch ('main') is prohibited
(ORCHESTRATOR §3) — main moves only via a merged PR.
.codearbiter/gate-events.log
[2026-08-03T07:59:05Z] BLOCK [H-01] host=claude hook=pre-bash.py$ git add -A
BLOCKED [H-03]: 'git add -A' / 'git add .' / 'git add --all' / 'git add -u'
are prohibited. Stage files explicitly (commit-gate skill).
.codearbiter/gate-events.log
[2026-07-30T13:34:06Z] BLOCK [H-03] host=claude hook=pre-bash.pyThese historical captures exercise the local Python hook at the pinned source revision below. Hash that file and compare. They do not qualify a newer package, host discovery, or an entire workflow.
Hooks mediate configured host tool calls. The evidence model is cooperative: unrestricted same-user filesystem access is outside its protection. See the security policy for the current product boundary.
The system
codeArbiter brings repository-owned context and governed development lanes to supported coding hosts. arbiterIDE is a separate editor project. Each product owns its implementation, threat boundary and qualification evidence; a shared design direction is not a shared release guarantee.
Repository-owned governance for Claude Code and Codex CLI, with Pi in preview. Commands route work to shared procedures and role-specific reviews; the product reference owns the current catalog and host exceptions.
Governed workflow requirements
Nine phases in the governed commit lane.
Tests, lint, secrets, behavioral proof and diff review are part of the configured commit procedure. Hook-backed blocks and model-guided requirements have different enforcement boundaries.
A reviewer fleet matched to the change, an 11-lens tribunal when it matters, and user-attributed architecture decision records. Overrides and auto-decisions all land in a durable audit trail.
Hook entry scripts use Python stdlib; structured artifacts also use a pinned native helper. Model-assisted reviews use your configured provider. Checked-in .codearbiter/ state survives switching hosts, while workflow capability remains host-specific.
A separate editor project exploring tool-dispatch enforcement. Visit the product site for its current development and availability information.
arbiteride.com ↗Install
Use the product-owned instructions for the current qualified installation channel. A source checkout is not automatically the released binary-bearing package.
What it writes
Repository context lives in .codearbiter/; integration and accounting state also exist in Git and user-global locations. Startup can contact your Git remote and GitHub. Optional tribunal feedback requires per-run consent. Read the privacy policy for destinations, storage and removal behavior.
License
codeArbiter is distributed under AGPLv3. Consult the product licensing and contribution policy for current terms and commercial availability.
How we work
You decide. The gates enforce.
Ask for the outcome. Keep the proof. The record outlives the session, the branch, and the host that produced it.
Read the essay: Why hard gates →