AI Agents

Operum provides 6 specialized AI agents, each designed for a specific aspect of building and growing a product. Choose the AI model and conversational style for each agent — customize how your team thinks and communicates. Communicate through PM or write directly to any agent. Add files, direct them to your folders, attach screenshots.

PM Agent

CEO / Product Owner

Your single point of contact. Understands your vision, breaks it into tasks, prioritizes work, and coordinates the entire team.

Architect Agent

Software Architect

Designs systems and chooses the right tech stack before code is written. Reviews scope, identifies edge cases, ensures solid foundations.

Engineer Agent

Software Engineer

Writes code, implements features, fixes bugs, and creates pull requests. Real commits, real branches, real PRs — all for your review.

Tester Agent

QA Engineer

Reviews and tests everything the Engineer builds. Catches issues before you see them — functional tests, edge cases, regression checks.

Marketing Agent

Growth Lead

Deep-researches competitors, target audience, and market opportunities. Creates content, manages SEO, and drives growth.

Community Agent

Support Manager

Handles customer support, engages users, and builds your community across Discord, Twitter, and more.

Coming Soon

More specialized agents are in development — see the Roadmap for details.

Coming Soon Accountant Agent

Financial tracking, invoicing, and expense management.

Coming Soon Sales Agent

Lead generation, outreach automation, and CRM integration.

Coming Soon Lawyer Agent

Contract review, compliance checks, and legal documentation.

Running Multiple Agents

You can run all agents simultaneously. They coordinate automatically — the PM delegates tasks, agents pick up work, update progress, and hand off to the next agent when done. Each code-touching agent (Engineer, Tester, Architect) works in its own isolated git worktree, so they can develop in parallel without conflicts.

Agent Settings

Each agent can be configured from the Agents page:

  • AI Model — Choose which Claude model powers each agent (e.g., Opus, Sonnet, Haiku)
  • Conversational Style — Professional, concise, creative, or casual
  • Custom Instructions — Additional per-agent prompts for specialized behavior

Starting & Stopping Agents

Agents don't run automatically — you control when they're active. Click Start on the Agents page to activate an agent, and Stop to pause it. Start with the PM first, then add other agents as needed.

What's Next?