Mission Control v2.1.0 is a self-hosted control plane for AI agent fleets - not another chat UI.
If you run more than one agent, you eventually need a place that answers three questions without opening six terminals:
- What is running?
- What is blocked?
- What can I approve or kill?
That is the product surface. The model is replaceable. The control plane is not.
Originally published as a launch note on X. This post is the long-form install-and-operate guide for the same release.
What Mission Control is for
Mission Control is open source under Builderz Labs: a dashboard template for multi-agent work. Think fleet status, job queues, and operator actions - not a closed SaaS that owns your loop.
You want it when:
- You have more than one agent profile or project
- Cron and human gates both exist in the same day
- You need a single pane for "what failed overnight"
You do not need it for a single Claude chat tab.
Repo: builderz-labs/mission-control
What v2.1.0 optimizes for
Release notes are boring unless they map to operator pain. This cut focuses on:
- Self-host first - you run the stack; credentials stay on your side of the network
- Agent visibility - jobs and sessions show up as objects, not log archaeology
- Handoff shape - a teammate can open the board and know what to do next
If your previous setup was "grep the agent log and hope," this is the upgrade.
Install shape (high level)
Exact commands live in the README. The mental model:
- Clone the template
- Configure env for your agent backends and auth boundary
- Run locally until the board reflects real jobs
- Deploy behind your own auth (never open a control plane to the internet without a gate)
Treat it like any other production dashboard: secrets, TLS, least privilege.
How it fits the nyk stack
On nyk.dev/oss Mission Control sits next to Hermes, Council, LACP, and xint. Different jobs:
| Surface | Job |
|---|---|
| Hermes / skills | Behavior contracts the agent loads |
| LACP | Local control plane / policy gates |
| Mission Control | Human-readable fleet board |
| Consulting | Install the loop with you (Agent OS package) |
Public code is the install path. Private ops stay private.
Operator checklist after install
- One real job visible end-to-end
- One failed job with a readable error
- One approve / reject path that a human can complete in under a minute
- One handoff note a teammate can follow without you on the call
If any of those are missing, you still have a demo.
What this is not
- Not a replacement for your agent host
- Not a closed multi-tenant product on this page
- Not "AI magic" - it is UI and structure around work you already run
Next steps
- Clone and read the README: mission-control
- Free pattern checklist: Agent OS setup checklist
- If you want the loop installed with you: Agent OS Setup package
Ship the board. Then ship the jobs that fill it.