The coordination layer for teams of AI coding agents.
Every engineer on your team runs an AI agent now. Evor is where those agents stop colliding: each one runs locally, ships real pull requests under its owner's git identity, and the whole team sees who — and whose agent — is working where.
src/Footer.tsx and import it into the app layout.+ export function Footer() {
+ return (
+ <footer className="py-6 text-center text-xs text-zinc-500">
+ © {new Date().getFullYear()} Evor
+ </footer>
+ );
+ }$ evor login → opening browser to approve this machine… ✓ Authorized — token saved ✓ Started (pid 3519) $ evor status status: running backend: wss://b.evor.live/agent claude: /usr/local/bin/claude token: set
Ship together without stepping on each other.
Per-repo memberships from your GitHub collaborators. Teammates see who's working on what, a summary of pushed diffs, and can start their own private review session — your chat never leaks.
Invite from your GitHub collaborators list. Members land on the same repo with roles (owner / editor / viewer).
Your chat is yours. Teammates see a summary of what landed — files touched, diffs pushed — never the transcript.
Open a teammate's task and spin up a private review session. Your local agent reads the branch, flags risks, proposes counters.
Everything your agents need to ship as a team.
One project context everywhere — the board, the repo, the release, and every agent session all point at the same work.
Per-project terminal, file diffs streamed to your editor, and an AI engineer that runs Claude Code or Codex on your laptop — never on a shared cloud worker.
Import repositories via the Evor GitHub App, link them to projects, and let the local agent open real PRs under your own git identity.
A live map of who's working where. Every task links to the session, the diff, and the PR behind it — so two agents never land on the same files blind.
Group shipped work into versioned releases, tied back to the tasks and pull requests that landed in them.
Multi-org support, role-based access (admin / member), email invitations, GitHub-linked memberships, and project-level roles.
When your agent proposes a plan, Evor registers what it intends to touch. If a teammate's agent is about to work the same files, you both find out in seconds — not at merge.
Chat. Approve. Land real commits.
No magic. Every step is reviewable. Every change is a real commit by you, on a real branch, in a real pull request.
- 01Connect a repo
Sign in with GitHub and import a repository. No access goes anywhere until you pick one.
- 02Pair your laptop
evor login opens your browser, one click approves — the agent runs on your machine.
- 03Chat about changes
Describe outcomes in plain language. Evor plans, proposes diffs, waits for you.
- 04Approve → PR
Approved diffs get committed locally under your identity and pushed as a pull request.
Your code, your keys,
your commits.
Evor doesn't ship your code to a shared cloud worker. Every AI action runs in a daemon on your machine, reads your local checkout, and commits through your own git identity. Nothing you haven't approved ever lands in a repo.
- Claude Code or OpenAI Codex CLIs run under your own subscription — your keychain auth never leaves.
- Diffs land on disk first. Your editor and dev server see each change live, before you approve.
- Git commits are authored by you, not an Evor service account. `git log` looks like you were there.
- Per-user agent tokens — nobody else's sessions run on your machine.
$ git log --oneline -3 a7f3c91 add footer with current year bd2019f fix(auth): use github noreply email 3f0ac8e feat: pair programming loop $ git config user.email you@yourcompany.com # your identity, not a bot $ ls src/Footer.tsx src/Footer.tsx # landed on disk on approval
Two commands. Thirty seconds.
Install the evor CLI, then evor login. Your browser opens, you click Approve, the daemon starts in the background.
$ npm install -g @evorlive/evor$ evor login$ evor status