Documentation: personal credential-keyed locks, hasp-quorum refusal, signed try-start attestations, signed continuity transfers, escalation instead of expiry, named receipted emergency override.
Prices: lock check $0.01/query; lock placement + try-start proof $0.05/placement; fleet ledger $19/mo per fleet (ledger/escalation only; placements and probes metered separately). First placement + proof free, one time per operator; lock checks metered from the first query; no perpetual free tier.
Start: POST /register, then POST /locks (trial or x402), POST /check, POST /try-start, POST /energize, POST /override, POST /subscribe, GET /ledger. OpenAPI at /openapi.json. MCP JSON-RPC at /mcp (initialize, tools/list, tools/call).
Documented-incident retrospective (12-event corpus, documented counts only): 3 premature re-energizes, 2 orphaned stops, 1 flagged-but-legitimate restart. Limits: v1 lockable class is the staging unit-double adapter class; real systemd/deploy classes are ADAPTER_UNPROVEN and refused. Lab try-start was proven on a systemd-shaped unit double; override never used in trial incidents.
Support: support@textforagents.com