About 05: the trust layer for agent-written software
Agents will write most code. What gets scarce is not typing but trust. 05 is the acceptance layer that lets people keep owning the intent and the outcome.
What 05 is
05 is an acceptance layer for AI-written code that takes the place of pull-request review. Humans own the request and the acceptance. Agents own the plan and the work, and 05 holds them to the plan, including several agents in one repository. The steps are on how 05 works.
Where it came from
05 comes from shipping real work with AI coding agents, on personal and client projects, through every generation of tools from the first ChatGPT to Claude Code and Codex.
The same problems kept showing up. AI is excellent at executing, but as a codebase grows, a great plan does not always give the outcome you expected. An agent can open more pull requests in an afternoon than a person can honestly review, so people either approve everything unread or let agents push with no review at all. Both end the same way: changes to the wrong files, tests quietly weakened, and systems nobody fully understands. Add a second or third agent and they start editing each other's files.
So the controls were built one at a time, as separate tools: one to state the intent before the work, one to check the outcome after it, one to keep several agents out of each other's way, and one to clean up the branches and worktrees they leave behind. Those tools have been in daily use for over a year, changing as the models and coding tools changed. 05 brings them together as one product, so that checking, approving, landing and pushing replaces pull-request review. A product for teams, 05 for teams, is on our roadmap and launching soon.
Why we are building it
Agents write code faster than people can review it, and pull-request review was built for people reviewing people. "Tests pass" from an agent is an assertion, not a fact. And prompting is an administrative control, not an engineering control. We think the answer is to move trust out of the agent's say-so and into independent, signed evidence attached to every change.
Principles
- Own the truth, rent the enforcement. Receipts, policy and identity are 05's and work on any host. Git hosts are thin, replaceable enforcement adapters.
- Honest about enforcement strength. Local hooks are advisory, and the CLI says so.
- Deterministic gates. No AI judges a gate.
- Stay on GitHub, not become a Git host. GitHub first, with a host-agnostic core.
- Any agent. Claude Code and Codex first.
- Your code stays yours. The local tool runs entirely on your machine. 05 for teams, launching soon, is designed to handle metadata only, never your code.
Where it stands
We are not showing customer logos, testimonials, user counts or pricing, because there is nothing honest to show yet. The core, meaning the local CLI and the portable runner, is open source under Apache-2.0, at github.com/use05-dev/use05. The numbers on this site are measured on the current build and are labelled that way.
Contact
The source is on GitHub. For anything else, email [email protected].
Keep reading
Try 05 on your own repositories
The source, install steps and docs are on GitHub. Questions, or want to tell us which agent and Git host you use? Email us. No forms, no cookies.