Your workspace,
in every AI tool.
Rezee is an MCP server. Connect the agent in your terminal, the assistant in your editor, a chat in your browser or an app on your desktop, and it works your issues, code, pull requests, docs, designs and chat as you.
~/code/web on kobeni/mask-env
❯ agent "What's left on REZ-42?"
● rezee.issue_get(workspace: "rezee", number: 42)
└ Encrypt pipeline secrets at rest · In Progress · kobeni
● rezee.issue_children(workspace: "rezee", number: 42)
└ 3 sub-issues: 1 done, 1 in progress, 1 todo
● rezee.pull_get(workspace: "rezee", repo: "web", number: 131)
└ Mask secret values in the runner env · open · checks running
● REZ-42 has one piece of work nobody has started:
✓ REZ-44 Per-workspace key derivation, merged in web#128
◐ REZ-45 Mask values in the runner's env, web#131 with tests running
○ REZ-46 Redact matches in streamed logs, no one on it
Start REZ-46 on a new branch? (y/n)
❯
Connect in a minute,
from the tool you use.
Most clients need only the URL and sign you in through the browser. The rest run the Rezee CLI locally. Either way, one connection reaches every workspace you belong to.
Claude Code
- 1Add the server from your terminal.
- 2Run
/mcpinside Claude Code and choose Rezee to sign in.
claude mcp add --transport http rezee https://api.rezee.app/api/mcp
Over 290 tools,
across every surface.
One tool for each thing worth doing, not one for each endpoint. They call the same API the web app uses, so your client can do what you can do, and nothing more.
Plan
File, triage and update issues, run cycles and projects, set custom fields, and hand an issue to an agent.
Code
Browse a repository, read files and diffs, open and review pull requests, and comment on a line.
Ship
Read pipeline runs and their logs, rerun a failed one, and report a deploy.
Design
Read and edit a canvas, render it to an image the model can see, and check it against its design system.
Docs
Read, draft and revise documents, answer review threads, and cut issues out of a spec.
Chat
Read channels and threads, post, react, and schedule a message for later.
Agents
See what your workspace's agents were asked to do and how it went, before starting the same work twice.
Your workspace
Your inbox, members and teams, and one search across issues, docs, chat, code and past agent tasks.
You decide,
what it may touch.
A connection acts as you, with only the permissions you ticked when you signed in. You can see and end every one from your account settings.
Sign in, don't paste
Give your client the URL and it sends you to a consent screen. No token to copy, and every connected app has a Disconnect in your settings.
Reads on, writes off
Reading arrives ticked. Writing issues, pull requests, docs or chat is offered unticked, so granting it is something you chose.
Some things it never gets
A connector cannot push with git, create repositories, see CI secrets or cut releases. Those cannot be granted.
Your client knows what is risky
Every tool says whether it only reads or cannot be undone, so a client that asks before acting asks at the right moments.
Asks for more, doesn't just fail
A call outside the grant names the permission that would have been enough, so your client can ask you to widen it.
Short-lived tokens
Access lasts an hour and is bound to this server. Codes are good for one use and one minute.
Your client, or
a teammate of its own.
The same tools serve two kinds of caller.
Your AI client wants to use your Rezee account
- Read code, issues, docs and chat
- Write issues
- Write pull requests
- Push code
A connector acts as you
Any MCP client signs in as you, and holds only the part of your permissions you ticked.
kobeni wants to merge
Waiting for your approval
Merge web#131 into main
pull_merge
An agent is a teammate
A Rezee agent has its own name and token, its work is credited to it, and it stops for a person before anything that cannot be undone.
Meet the agentsBring your next project.
Start free, with no card. Connect the agent you already use, hand it an issue, and approve its first merge today.