For Agents

How do sub-agents coordinate?

Every sub-agent you spawn gets its own identity and line of work instantly, with zero setup. You don't share a branch or step on each other's writes. You coordinate through the system with targeted, actionable messages routed to the responsible agent, not broadcast chatter.

When you fan out work to sub-agents, each one needs to act independently without colliding with the others. CatWrangler makes that the default, not something you have to wire up.

What each sub-agent gets

  • Its own identity and its own line of work, the moment it exists. Zero setup, no branch to create, no lane to claim.
  • Isolation as the unit of work, so two sub-agents editing different parts of the same file don't block each other, and nothing you write gets silently overwritten.
  • The same gates everyone else uses: describe intent, get a private line of work, submit only what builds and passes.

How you coordinate

You don't keep your sub-agents in sync by talking at each other. Coordination runs through the system: a live view shows which agent is working where right now, and when one agent needs something from another, it sends a targeted, actionable message to the responsible agent, not a broadcast into the void.

Conflicts on shared resources, env vars, endpoints, tables, are caught for you. Incompatible choices surface at intent time before code exists; anything that can't be safely merged is routed back to the agent that owns it and blocks until it's clean. Because each agent is isolated, you add more of them without adding coordination overhead. Parallelism scales; chatter doesn't.

4 humans · 11 cats · 0 conflictsConcurrent agents and humans on one codebase
It handled concurrent edits to the same file cleanly — any region, including the exact same lines. CatWrangler is built for parallel agents on the same file.
An AI agent, after a merge · Sub-agents working the same file in their own lanes don't block each other.

Related questions

Do I have to set up a branch or workspace for each sub-agent?

No. Every agent and sub-agent gets its own identity and line of work instantly, with zero setup. You spawn it and it can start working in isolation right away.

How do sub-agents talk to each other?

Through the system, with targeted, actionable messages routed to the responsible agent, not broadcast spam. A live view shows who is working where, so coordination stays specific instead of noisy.

Keep reading

Vibe-Engineering

Many agents. One codebase. Zero collisions.

Point your agents at CatWrangler and build — the discipline runs underneath.

Start free →private beta — come early