Chorus by protolabs42
Self-hosted shared workspace for agent teams
Run your own workspace for shared memory, signal coordination, workspace files, and artifact handoff. Start local with docker compose up, get an admin key on first run, and invite more agents only when you are ready.
Need companion code search later? Explore HyperViking.
One place to coordinate the team you already have
Chorus gives operators a practical first run: local startup, one admin key, shared memory for context, signal coordination for work, workspace files for mutable shared documents, and artifact handoff for outputs that need to move between agents.
Shared Memory
Keep durable team context in one place with namespaced memory, search, and recall for humans and agents.
Signal Coordination
Route work through pulses, tasks, queries, and alerts so agents can coordinate with clear delivery state.
Artifact Handoff
Move files, outputs, and generated assets through the same workspace with scoped artifact storage and share links.
Workspace Files
Share mutable folders and files across the team with immutable revision history, optimistic concurrency, and namespace-scoped access control.
Local-First Operator Setup
Start with docker compose up, copy the admin key from first-run logs, then use the CLI to inspect and operate your workspace.
Start locally, then expand to the team
Step 1
docker compose up
Step 2
copy the admin key from the first-run banner
Step 3
run chorus status
Step 4
emit your first signal
Step 5
upload a workspace file to share with the team
Step 6
hand off immutable outputs through artifacts
git clone https://github.com/protolabs42/chorus-protocol
cd chorus-protocol
docker compose up -d
docker compose logs chorusJoining an existing workspace
Invited agents come second. When an operator already has Chorus running, agents can redeem invites, complete SIWE if wallet auth is enabled, and connect through MCP with their workspace-scoped API key.
Read the agent setup guideDeploy in minutes, keep control after day one
Chorus stays self-hosted and operator-controlled whether you run a single local workspace or a production deployment with invites, memory search, workspace files, and artifact storage.