Sterile Cockpit for Irreversible Actions — API docs (TextForAgents LLC) Base: https://textforagents.com/products/sterile-cockpit-for-irreversible-actions/ Auth: POST /register (name, wallet) returns a bearer token shown once. Every phase verb requires the owning operator's token. POST /phase/open {effect_class, intent:{counterparty, amount, object}} -> 201 {phase_id, intent_hash, price_usd, rail} | 402 x402 challenge $0.02 after the one free authenticated phase. POST /phase/{id}/input {cls, payload} -> 202 deferred. Classes: user_message, tool_refresh, memory_write, memory_retrieval, subagent_return, compaction, model_switch. Deferred content hard-bounded at 1 MiB per phase. POST /phase/{id}/break {kind: ABORT|SAFETY, source, reason} -> 200; a break VOIDS the frozen intent — no resume path exists. POST /phase/{id}/effect {effect} -> 200 frozen signed exit receipt envelope; an effect diverging from the frozen intent is refused 409 and counted. GET /phase/{id}/receipt -> the operator's own frozen receipt (free read path). POST /verify {receipt} -> counterparty verification, $0.01 via x402 (402 challenge without payment). POST /subscribe -> Fleet plan $19.99/mo via x402: exactly 2,000 phases/mo included; past-cap phases meter $0.02. GET /health, /healthz — liveness only. Demurrage table (free window per effect class, then $0.001/second, bound into the exit receipt): { "delete": { "free_window_s": 60, "rate_per_s_usd": 0.001 }, "deploy": { "free_window_s": 300, "rate_per_s_usd": 0.001 }, "order": { "free_window_s": 30, "rate_per_s_usd": 0.001 }, "pay": { "free_window_s": 30, "rate_per_s_usd": 0.001 }, "publish": { "free_window_s": 60, "rate_per_s_usd": 0.001 }, "send": { "free_window_s": 30, "rate_per_s_usd": 0.001 } } Stated plainly, because the lab proved the boundaries: the lab's divergence result (0/30 protected vs 30/30 unprotected) is a sandbox injection result; how often mid-phase inputs land in real production windows is UNMEASURED and no frequency, incidence, or error-share figure is stated anywhere; sterility proves the executed effect is the frozen effect — it does not certify the decision itself was wise or policy-authorized (sterility is not decision correctness); the free phase is for authenticated operators only. Support: support@textforagents.com