Closeout has no receipt
Your client asks "what ran?" and you forward logs. They wanted a signed PASS or BLOCK — something counsel can file.
The problem you already feel · canonical platform page
Single source for platform buyers — API hook, trust strip, procurement links. Home (/sourcea/) is the integration hub; this page is the evaluation story.
Policy at dispatch — one call, one receipt
POST /v1/decision
→ {"verdict": "BLOCK", "receipt_id": "sa-0886", "signed_at": "…"}
// Live from factory disk · refreshes on deploy
Same spine as sourcea-boot — eval the gate before you buy the build.
Execution Proof Infrastructure
Your client asks "what ran?" and you forward logs. They wanted a signed PASS or BLOCK — something counsel can file.
Governance at session start misses the dangerous action at minute forty. You need a gate on every dispatch.
Workflow tools connect steps. They do not prove outcomes, catch tampering, or let you replay yesterday cold.
Our factory runs these loops nightly — you inherit the spine we dogfood.
Every action passes your rules on disk before the LLM executes.
Weekly export with what ran, what blocked, and which policy applied.
ALLOW, BLOCK, and tamper detection — screen-shared in under five minutes.
Before you read another slide deck
Receipt counter from disk · GitHub eval path · downloadable proof bundle · governance status page.
Built on
Dependencies we integrate — not paid partner badges · Makes Claude deployable with policy gates + signed receipts · API /api/runtime-wrapper/v1 on Hub
How your loop stays governed
Each proposed action checked against your policy — every time.
Sub-millisecond checksALLOW or BLOCK on disk — Monday starts from evidence, not memory.
Synced with your rosterRe-run last night without chat history. Tamper shows up immediately.
Demo-ready in minutesLoops we run for clients
Go deeper
Walk the chain from request to receipt — live, in the time it takes to finish a coffee.