Development runs
From an approved issue to a merged pull request
A development run hands an approved, well-specified issue to a coding agent. The agent works on a branch and opens a pull request. Your team stays in charge of review and merge.
The run's contract
Only approved issues
An issue becomes a candidate only when it carries the approved label. One run per issue at a time.
Started your way
An operator starts the run, or the project starts runs automatically on approval.
A branch, tests, a pull request
The agent works on its own branch, implements the change with tests and opens a pull request that names the issue.
Stops at merge
The run answers review comments and failing checks, and stops at merge. It never pushes to a protected branch.
Followed from start to release
Every state recorded
Queued, running, pull request open, changes requested, CI red, green, merged: every change is on the run's timeline.
Woken when needed
When a reviewer asks for changes or CI turns red, the agent is woken to answer. A person can take over at any time.
Stalls are flagged
A run that stops making progress alerts the operators, who resume it or abandon it. The branch is kept for a person.
Linked to the release
Your deploy reports the release it shipped, so Maestro shows which release carried which run, and the product loop announces it.
Runtimes and tooling
Claude Code
Runs use Claude Code through a GitHub Action in the project's repository, switched on per project from the onboarding checklist.
TALOS MaiND development agent
A TALOS MaiND agent with repository tooling, as a second runtime chosen per project.
A GitHub App per repository
Maestro reaches GitHub through a GitHub App installed on each repository, with access tokens created for each call.
Driven from your own sessions
Maestro includes an MCP server, so a developer's own Claude session can check a project's status and drive its runs with a personal, expiring token.
Notifications
In the console and by e-mail
A run opened a pull request, needs a person, merged or stalled; an agent went stale or a publish failed. Operators hear about it.
Bring your product team into the loop
Maestro is ready to deploy on your product. Tell us about your product and your team, and we will show you how Maestro runs it.
Your product's AI team and delivery pipeline, with people at every gate.
Maestro is part of the TALOS MaiND family.