AI agents that work on your issues
Start an agent from an issue, in the cloud or on your own computer. It reads the issue and the knowledge base, writes the change in a branch and opens a merge request under your account.
In the cloud
A fresh machine per issue with Claude Code or Codex, on the AI credit of your plan or on your own API key. The agent sees the project through read-only MCP tools and the knowledge base of that project. The machine is gone when the work is done, and a key of yours never lands on it.
On your computer
Pick a computer in the same panel: this one or another you paired with the desktop app. The agent starts there, in a local clone, with your own subscription or API key. Code4 does not see your account with the model provider and only keeps the record of the session.
As you, under your permissions
The branch, the push and the merge request go under your account. Protect the default branch, and the change reaches it only through review. The in-app assistant asks before actions that are hard to undo, such as merging or deploying.
The same context as your team
Agents answer from the same knowledge base as people: issues, reviews, documents, chat and call transcripts, under the permissions of whoever started them.
Rules and skills, for the organization and for each project
Write how AI should work in your organization once, in its settings, and add what is specific to a project in that project. Conventions, review rules, how to run and test, step-by-step skills for recurring work: every agent and the in-app assistant gets them over MCP, whichever tool it runs in.
From an issue to a merge request
- 1
Open the issue and click Agent
One panel, one choice: a fresh machine in the cloud, or your own computer with the desktop app.
- 2
The agent reads the issue
Through MCP, as you: the issue, the linked code and the knowledge base of the project.
- 3
It writes the change in a branch
A branch named after the issue key, a failing test first, then the change.
- 4
It opens a merge request
The issue key in the title links the request to the issue; your reviewers take it from there.
- 5
The record stays on the issue
Status, branch, merge request and a short summary of what was done.
Questions about agents
- Which agents can we use?
- In the cloud: Claude Code or Codex, on the AI credit of your plan or on your own key for Anthropic or OpenAI. On your computer: Claude Code through the desktop app, with your own subscription. Cursor or any other MCP client connects to the same server and gets the same tools.
- Does an agent need its own account or seat?
- No. It acts as the person who started it and sees exactly what that person sees. Agents do not take a seat on your plan.
- Can it merge or deploy on its own?
- Over MCP an agent has the rights of the person who signed it in, so protect the default branch and keep deploy approvals on. The in-app assistant asks before actions that are hard to undo.
- Which plans include agents?
- Cloud agents are on every plan, including Free: on the AI credit of the plan, or on your own key. Starting an agent on your own computer from an issue is on Premium and Business.