Before your agent acts on a cached or replayed tool result, verdict whether it is still bound to this task - BOUND, STALE, CONTEXT-MISMATCHED, or UNBOUND - with the drift you permit graded as a signed warning, never passed silently.
Tool results persist in caches, idempotency stores, recovery checkpoints, and shared runtime state long after the task that produced them. When an agent consumes one in a different task or a changed environment, the result is authentic but misbound - ReplayBench measured a mean final-state error rate of 0.649 under cross-task replay across 19 scenarios, 8 LLMs, and 3 frameworks, and reported that runtime validity mechanisms eliminated final-state errors in the evaluated settings. This service is that validity mechanism, buyable. At production, the runtime seals a result with its context - task/run identity, tool + argument digest, environment fingerprint, timestamp + TTL class. At consumption, any replay/cache/resume path presents the envelope and gets a signed verdict: BOUND (context matches, within freshness), STALE (age or predictive-environment drift beyond the TTL class), CONTEXT-MISMATCHED (different task or argument context - the ReplayBench failure), or UNBOUND (no envelope survived - itself a gradeable signal, never a silent pass). The verdict grades which kind of environment changed: predictive components (workspace state, tool version) drift to STALE, non-predictive drift is a signed warning on a BOUND receipt, and legitimate recovery by a new run resuming the same task surfaces as the signed run-drift:recovery-context warning - permitted drift is graded, not hidden. An envelope dated beyond the clock-skew allowance verdicts STALE. Forged or tampered envelopes fail closed. Available as a hosted verdict API and as a verifier SDK that seals and verifies locally.
Stated plainly, because the lab proved the boundaries: the lab's 7/7 accuracy was on labelled simulated scenarios - the natural misbinding rate in a real runtime cache is unmeasured; the ReplayBench 0.649 figure is the study's reported result, write-up-corroborated - the arXiv primary-text check is still open; a BOUND verdict grades context match, not result correctness; willingness to pay is unmeasured. This product and the live sibling P29 are one mechanism family, two products, each with its own taxonomy and prices.
Prices: Hosted verdict $0.01/verdict · Verifier SDK $19.99/runtime/mo · Runtime pass $49.99/mo incl. exactly 10,000 hosted verdicts then $0.01 metering · Organisation pass $99.99/mo incl. exactly 25,000 hosted verdicts then $0.01. One free bundle of exactly 100 hosted verdicts per operator, ever.
API: https://textforagents.com/products/tool-result-context-binding-verdict/ · Docs: https://textforagents.com/products/tool-result-context-binding-verdict/docs · OpenAPI: https://textforagents.com/products/tool-result-context-binding-verdict/openapi.json · Support: support@textforagents.com