gitring

gitring calls itself source control for coding agents: when several agents work in one repository, it does not wait for merge conflicts — agents must declare intent before they write code.

In one call, an agent says which files and symbols it will touch and why. gitring scores how close two plans are, applies an ordered rule set so one agent proceeds and the other adapts, and asks a human only when the rules cannot settle it. Agents talk to it over MCP; people get a web UI.

Highlights:

• Review the intent, not a wasted diff;
• Built for “one repo, many agents,” not another Git host;
• In a published load test, 60 agents at once declared intent at ~69–144 ms p95 and 1,512/1,512 merges landed on main;
• Closed source (Mindalike); public since 2026-11-03; sign in with GitHub or Google; site https://gitring.tech , app https://app.gitring.tech .

Why it matters

Labs such as Anthropic have described merge floods from coding agents and built internal systems (e.g. Scion). Most teams cannot rewrite source control, yet already run several Claude Code / Codex agents on one repo. gitring moves conflict handling to the planning step — a lighter coordination layer than a custom VCS.

Limits

The product is closed source. It only works if agents actually declare intent and if the rules match how your repo is structured. It does not replace code review, tests, or CI; it only answers “who may touch what next.”

1. Open https://gitring.tech or https://app.gitring.tech and sign in with GitHub or Google;
2. Connect the repository and have coding agents declare intent (files/symbols + why) over MCP before they edit;
3. Use the web UI to watch live intents, overlaps, and the queue; approve only what the rules cannot settle;
4. Agents edit according to the schedule; gitring can check the landed work against the plan.