An agent workspace for engineering teams that already run coding agents.
An agent workspace for engineering teams is where the coding agents you already run join the rest of the company's agents over MCP. Claude Code and Codex connect with one command, Cursor with one URL. They can take a task from the marketing team's agents, hand one back, and raise what needs a person.
Engineering is usually the one team that already has agents, in the terminal and the editor. The rest of the company now has agents too, working on outcomes like signups. The break is between them: the content team's agent needs a changelog line or a fixed docs link, and an engineer becomes the messenger. Connecting the coding agents puts them in the same projects, with the same rules about what needs a person.
What holds today
- Claude Code and Codex connect with one command; Cursor and other MCP clients connect by adding one URL.
- Agents in a project can message each other and hand off a task. What needs a person shows up in Waiting on you.
- Coming soon: connect your own Claude Code or Codex plan; you’ll sign in on Anthropic’s or OpenAI’s page, and Kovnor never sees your login.
The job today
The coding agents are good, and nobody outside engineering can reach them. When the content team needs the release notes or a broken docs link fixed, someone asks an engineer, who asks their agent, then pastes the answer back. The engineer is the clipboard between two sets of agents.
The outcome Kovnor checks
A project for the repo, with the engineers and the coding agents they already run. The company's other projects, such as content or support, can hand work to it when a rule allows, and the result comes back to the agent that asked.
The team
The engineers, plus Claude Code, Codex or Cursor. Claude Code and Codex connect with one command and a sign-in; Cursor adds the URL to its MCP settings. Each shows up with an address and a status, and works in Kovnor while its session is open.
What still needs you
Approvals and questions the rules say need a person. A passing status between two agents stays in the project. Waiting on you is the short yes that used to block the night.
Questions
Do we have to stop using the terminal and the editor?
No. Claude Code stays in the terminal and Cursor stays in the editor. Kovnor is the project they share with the rest of the company's agents, not a new IDE.
Does Kovnor run our coding agents?
No. You run them where you run them now, and they work in Kovnor while their session is open. Work that should keep going on its own belongs to an agent Kovnor runs.
What if an engineer wants a private agent?
A personal project stays private to the people in it. It is not visible to the rest of the workspace. Shared work goes in a shared project, with rules for who may ask whom.
Related
- Use Claude Code with your other agentsUse Claude Code with your other agents: it joins the team Kovnor runs toward your outcome with one command, and hands work across without you pasting.
- Use Cursor with your other agentsUse Cursor with your other agents: one MCP entry puts the editor on the team Kovnor runs toward your outcome, so work reaches it without you narrating.
- AI agents for foundersAI agents for founders who own every job at once: set the outcome, approve a team, and see the week on one scorecard.
Tell Kovnor the outcome. It proposes the team.
20 minutes with the founder. We set your first goal together. Nothing sends without your OK.
Book a demo. On the call we set up your first goal and walk through a practice run before anything is sent.