Search hypothesis · private beta
Reconstruct what an agent did—and what it did not do
Inspect the run, provider/model identity when available, artifact, route, gate, correlation, and side-effect boundary without inferring missing proof.
Private beta. No autonomous publishing or production mutation without approval.
Product-true proof
Missing receipt fields stay missing
Launchfiles distinguishes request acceptance, artifact verification, lifecycle execution, and approved side effects. One receipt never silently stands in for another.
- Run
- Durable run or invocation identifier
- Artifact
- Identifier and route-specific output
- Gate
- Pass, debt, or honestly blocked state
- Boundary
- Observed side effects and unavailable correlation
Mission handoff
From problem to next safe action
- 01
Bound the mission
Name one decision, the evidence available, the owner, and actions that remain forbidden.
- 02
Return the artifact
Produce the route-specific brief, work orders, gates, and honest evidence gaps—not a transcript.
- 03
Expose the boundary
Show what ran, what did not run, what needs approval, and the next safe action.
30-day evidence ceiling
This page must earn its place.
Rename this page if “agent receipts” is interpreted mainly as billing; retain the auditability job and test clearer language.
Contamination boundary: customer-product capabilities may appear only as labeled case-study evidence, never as Launchfiles features.
Private beta
Bring one founder mission.
Leave with a useful artifact, evidence, an approval boundary, and the next safe action.