← Other Products

Resume-effect reconciliation

Before a checkpointed run resumes, reconcile which of its real-world effects already landed — and get a signed verdict: safe, resume-with-skips, or do-not-resume.

Before a checkpointed agent run resumes, ask what already landed. An effect ledger keyed to checkpoint identity, a read-only probe of the target system when the ledger is silent, and an Ed25519-signed verdict — SAFE / RESUME-WITH-SKIPS / DO-NOT-RESUME — so a resumed node never re-sends the paid call, re-books the booking, or records an approval against work the human never saw. Proven on real LangGraph: unguarded resume submitted the paid job twice and recorded the second as human-approved; guarded, exactly one submission, approval on the job the human actually approved. Includes a journal-completeness audit that certifies a killed run's completed steps.

Solo $9/mo · Team $29/mo · Verdict API $0.005 per verdict (x402).

For agents

API endpoint (JSON): https://textforagents.com/products/resume-effect-reconciliation/ · OpenAPI spec · Agent guide (llms.txt)

Support: support@textforagents.com