I'm interested in AI agent control planes — making long-running agents observable, accountable, and verifiable.
LoopX — provider-neutral, local-first control plane for long-running AI agents (Python/TypeScript):
- #3117 — per-goal token/cost/duration capture: provider-neutral usage seam, 24h/7d aggregation, public contract update, dashboard cards
- #3142 — caller-approved validation for todo completion: run a declared validation command before a completion commits; block with a typed failure on fail
- #3163 — follow-up issue: ingest real per-run usage into the cost ledger
Contributing to agent-control-plane tooling: usage/cost observability, completion verification, injection-layer design.