For engineering teams
AI coding agents your whole team can watch
Evriq runs Claude Code, Codex and other coding agents in a session your whole engineering team can open. Agents change code in your repositories while people watch the work, step in, review the change with several models and ship it from the same room.
Add any repository by URL
Create an Engineering project and add its repository by URL. Evriq works with GitHub, including GitHub Enterprise, GitLab, Bitbucket, Gitea, Forgejo, Codeberg, your own server, or anywhere else Git lives. You also choose which host new repositories are created on.
When a clone needs a token, the credential goes to git through a prompt helper and is kept in the workspace vault. It is never written into the remote URL or .git/config, and clone errors are scrubbed of it before you see them.
Each session in its own working copy
Every session gets its own working copy of the repository on its own branch. Two sessions can work on the same codebase at the same time without changing each other's files, and two teammates can each run an agent on a different task without coordinating first.
Seats on Claude and Codex run as real coding agents that read files, edit them, run commands and commit. You pick the model for each seat in the model bar, and every reply says which model produced it.
Branch, diff, commit and push from the room
Engineering and General projects show the git tools. From the session you can see how far the checkout is from origin, read the diff, commit, pull and push, and merge a session's branch back.
If you commit without a message, Evriq writes a real subject from the changes. A push that would publish a secret value is refused, and the message names the secret, the file and the line. Every commit, pull and push is written to the audit log with its SHA.
Anyone on the project can follow along as the agent works. Take over makes you the lead when you want to redirect it, and Pass hands the session to a teammate with a note.
Several models review the change
Start a review round on a change and every seat reviews it blind, without seeing the others' answers. Each then sees the others once and may revise, and one seat writes the verdict, approve, request changes or reject, with reasons.
Having Claude review what Codex wrote, or the other way round, catches things a single model misses. The verdict is recorded on the session, and a work item for the repo change can go to a named reviewer for approval before it is marked done.
Run it from the terminal
The evriq command-line tool is a full client, and every task in the web app has a command. evriq session attach opens a live session in your terminal, so you can follow an agent's turn from the shell you already have open.
On a server you reach over ssh, sign in with the device flow and approve the code from any browser. The CLI help covers install and sign-in. For running several agents on one piece of work, see multi-agent orchestration.