Before I run a new local agent, I want a boring checklist.
Not hype. Not a demo video. Just operational basics.
- What will it install?
- Where will it store state?
- What provider credentials does it need?
- Which folders can it read or write?
- Which tools or MCP servers can it call?
- Does it run in a container or directly on the host?
- Where are the logs?
- How do I stop it?
- How do I resume a failed run?
- How do I remove it cleanly?
This is the layer I think local agents are missing.
Frameworks help you build agents. Model providers help you run inference. MCP helps tools plug in.
But operators still need a local control plane for setup, jobs, logs, approvals, and recovery.
That is what Armorer is trying to become:
https://github.com/ArmorerLabs/Armorer
And for consequential actions, Armorer Guard is the companion layer for decision receipts:
https://github.com/ArmorerLabs/Armorer-Guard
If you run local agents, I would love feedback on what belongs in this checklist.
Top comments (0)