# OpenGantry

> Deterministic CI/CD and governance for autonomous agents. Local CLI that scopes agent edits, runs your shell gates, and returns structured JSON failures.

OpenGantry is the `gantry` CLI: run `gantry legislate` to write a Mission YAML (gap interrogation is built into legislation, not a separate stage), execute in scoped paths across agentic graphs or single IDE sessions, trace evidence in Git, and verify with `gantry verify` at each gate. It is **not** [Gantry.io](https://gantry.io) — there is no hosted observability dashboard; everything runs in your repo. It works with orchestration buses (e.g. iii.dev middleware + verify workers), any agent harness (MCP + hooks), CI pipeline graphs, or human reviewers. The same mission contract governs code, documentation, and policy files.

## Pages

- [Home](https://opengantry.ai/): Deterministic CI/CD and governance for autonomous agents — scoped edits, shell gates, structured failures.
- [Features](https://opengantry.ai/features/): Discovery, blueprint, perimeter checks, mission verify, and harness hooks.
- [Cage](https://opengantry.ai/features/#cage): What `gantry cage` restores, refuses and halts, its limits, and per-project `.cage.yaml` rules.
- [Manifesto](https://opengantry.ai/docs/manifesto/): North Star positioning — outer-loop trap vs inner-loop referee.
- [Audit trail](https://opengantry.ai/audit/): Git-native mission evidence for security and compliance reviewers.
- [Get started](https://opengantry.ai/get-started/): Install the CLI, bootstrap a repo, and run your first mission.
- [Documentation](https://opengantry.ai/docs/): OpenGantry docs mirrored from the upstream repository.

## Product

- [OpenGantry on GitHub](https://github.com/jeger-ai/opengantry): CLI source, substrate, and examples.
- [npm package](https://www.npmjs.com/package/@jeger-ai/opengantry): Install with `npm install -g @jeger-ai/opengantry`.
- [Cage package](https://www.npmjs.com/package/@jeger-ai/opengantry-cage): Run any agent inside the cage without installing the full CLI: `npx @jeger-ai/opengantry-cage -- claude`.
- [Zero-config cage](https://opengantry.ai/docs/features/#zero-config-cage-gantry-cage): `gantry cage -- <agent>` restores CI workflows, `.env` files, git hooks and config while an agent session runs and refuses pushes of them; `.cage.yaml` adds your own paths (additive only), `gantry cage suggest` proposes them, `--allow-override <path>` relaxes one path for one run, always announced.
- [Adoption guide](https://opengantry.ai/docs/adoption/): Five-minute adoption walkthrough.
- [Integrations](https://opengantry.ai/docs/integrations/): MCP tools, hooks, and per-harness setup.
- [Agent graph](https://opengantry.ai/docs/agent-graph/): How external executors close the autonomous graph (verify gates + structured retry edges).
- [iii integration example](https://github.com/jeger-ai/opengantry/tree/main/examples/iii-integration): Orchestration bus — verify, middleware, RBAC hooks.

## Optional

- [Marketing site repository](https://github.com/jeger-ai/opengantry.ai): Astro source for opengantry.ai.
- [SECURITY.md](https://github.com/jeger-ai/opengantry/blob/main/SECURITY.md): Security policy for the OpenGantry CLI.
- [DOMAINS.md](https://opengantry.ai/docs/domains/): Built-in code and content domain adapters.