Beta mumus is in public beta: free while we test it with real teams. Found something odd? Tell us.

Automate your first-level support, with a person a click away

Most tickets need the same first steps: read the report, look at the screenshot, ask what is missing, check the code. mumus gives your boards an AI first-level support agent that does exactly that through Claude Code or Codex, and hands anything it cannot solve safely to the right person with a short internal briefing.

  • Answers questions, asks reporters for missing details and fixes simple bugs as pull requests
  • Hands off to a person with the reason, its findings and the next step, visible to your team only
  • Always labelled as AI; it never signs in, never merges, never deletes and takes no seat

What the AI agent does

  1. Picks up new tasks on the boards you choose.
  2. Reads the report, the screenshot, the console errors and failed requests that mumus captured.
  3. Answers questions it can answer from your code or documentation, and asks the reporter when details are missing.
  4. Fixes bugs it can fix safely: a branch, the fix, the tests, a pull request, and the card moves to Review.
  5. Hands everything else to a person.

Hand-off to a person, with a briefing

When a task needs access, a business decision, design work or a risky change, or when the customer simply wants a person, the agent does not guess. It assigns the task to the project's hand-off person, tags it "needs human" and writes an internal note that only your team sees: why it stopped, what it checked and what to do next. If you want, it also tells the customer that a colleague is on it.

Your code stays with you

The agent runs where your code is: on a developer machine or in your own CI pipeline, using Claude Code, Codex or another MCP client you choose. mumus never needs access to your repository and does not send your tickets to an AI provider itself. The agent only reads what its token allows, and every pull request is reviewed by a person before it is merged.

Clearly an AI, with clear limits

  • Shown with a robot avatar and an "AI" label on every comment, change and assignment
  • Cannot sign in, gets no e-mail and does not count as a team seat
  • Works only on the projects you choose, with Contributor rights; it never deletes or archives anything
  • One click revokes its token; its history stays

Track the human time

Time tracking shows how much human work each ticket really needed. You decide per board whether clients see the total or every entry.

Frequently asked questions

Does mumus run the AI itself?

No. The agent is Claude Code, Codex or another MCP client that runs on your machine or in your CI. mumus provides the board, the permissions and the hand-off workflow, and never sends your tickets to an AI provider on its own.

What happens when the AI cannot solve a ticket?

It hands the ticket to a person: the project's hand-off person or someone it names. The person gets the assignment, a "needs human" tag and an internal note with the reason, the findings and the suggested next step.

Can customers tell that they are talking to an AI?

Yes. The agent always has a robot avatar and an "AI" label next to its name, in comments, the history and the assignee list.

Does the AI agent cost a seat?

No. AI agents never count towards your member limit. Freelancer and Small team include one agent, Agency three, and Enterprise any number.

More questions in the help centre

Turn website feedback into tasks your team can fix.