mirror of https://github.com/different-ai/openwork synced 2026-05-14 19:16:24 +02:00

Files

ben 366bc36d55 feat(agent-lab): multi-instance host + Agent Lab Toy UI (#504 )

* feat(agent-lab): add instance manager CLI

* fix(server): block opencode permission replies for non-owners

* feat(toy-ui): add share + checkpoints

* feat(agent-lab): add entrypoints and scheduler sync

* feat(server): add agentlab automations and skill delete

* feat(toy-ui): add agent lab panels

2026-02-08 13:58:50 -08:00

src

feat(agent-lab): multi-instance host + Agent Lab Toy UI (#504 )

2026-02-08 13:58:50 -08:00

package.json

feat(agent-lab): multi-instance host + Agent Lab Toy UI (#504 )

2026-02-08 13:58:50 -08:00

README.md

feat(agent-lab): multi-instance host + Agent Lab Toy UI (#504 )

2026-02-08 13:58:50 -08:00

tsconfig.json

feat(agent-lab): multi-instance host + Agent Lab Toy UI (#504 )

2026-02-08 13:58:50 -08:00

README.md

OpenWork Agent Lab (toy)

This package is a CLI-first instance manager for the Agent Lab concept.

It is intentionally small and experimental.

Agent Lab reuses existing OpenWork infrastructure:

openwrk orchestrator (sandboxing + sidecars)
openwork-server edge API + Toy UI
opencode engine

Local development

From the repo root:

pnpm -C packages/agent-lab dev -- --help

Commands (MVP)

# Create an instance directory + workspace
pnpm -C packages/agent-lab dev -- create --name Scout

# Start services (sandbox) for that instance
pnpm -C packages/agent-lab dev -- start <instanceId>

# Open the Toy UI for that instance
pnpm -C packages/agent-lab dev -- open <instanceId>

# Stop the sandbox container
pnpm -C packages/agent-lab dev -- stop <instanceId>

Notes:

This is macOS-first and assumes Bun is available.
Multi-instance is achieved by per-instance ports + per-instance directories.