Build focused agents powered by your APIs and integrations. Control exactly what they can do. Safe for you, your team, and AI systems like Claude Code.
Join n+1 developers on the waitlist
Found 3 open PRs in acme/frontend:
#142 Add dark mode toggle (ready for review)
#139 Fix auth redirect bug (2 approvals)
#134 Update dependencies (draft)
> "Delete the frontend repo"
⚠ Agent acme/github cannot delete repositories.
Allowed: list_prs, list_issues, create_issue
Define once what an agent can do. Enforced every time it runs.
Build agents that do one thing well. A deployer, a notifier, a database querier.
Choose which APIs, which orgs, which repos. Define the boundaries - the agent can't exceed them.
Use it yourself, share with your team, or let Claude Code call it. Same guardrails apply.
Your agents, your rules. Enforced for humans and AI alike.
Define what an agent can do. Those limits apply whether you, your team, or Claude Code runs it.
Restrict agents to specific orgs, repos, channels, or resources. Granular control over what they can touch.
Update permissions in one place. Changes apply immediately to all users and AI systems.
Run agents from your terminal with moi workspace/agent. Works anywhere.
Let Claude Code, Cursor, and other AI systems call your agents as subagents. They get capabilities without direct API access. You keep control.
Focused agents with clear boundaries. Useful for you and safe for AI.
moi deployTrigger deploys to staging only. Production requires approval.moi githubList PRs and create issues in your org. No repo deletion.moi db-queryRead-only access to analytics tables. No PII exposure.moi slackPost to #engineering. Can't DM or access private channels.moi ticketCreate tickets in your project. Can't modify sprints or users.moi logsFetch logs from staging. Production logs require auth.Create CLI agents you can trust AI with. Join the waitlist.