The trust layer for autonomous systems.
Generate verifiable execution receipts for AI systems.
Logs tell you what an agent said. Receipts prove what it did.
The first thing you write is the receipt.
Wrap an AI call, agent step, or workflow and get back the original result plus a verifiable receipt.
// generate a receipt around any AI execution import { Memora } from "@smritheon/memora-core"; const memora = new Memora({ agentId: "treasury-agent" }); const { result, receipt } = await memora.receipt("ai.call", async () => { return await agent.run(input); }); // result is unchanged // receipt.execution_id is ready to verify
Logs tell you what agents said. Receipts prove what they did.
A log line is self-reported and easy to rewrite. A receipt captures the execution boundary: the agent, inputs, outputs, result, and verification state.
Editable · self-reported · unprovable.
Structured · verifiable · replay-ready.
Receipt → Verify → Replay.
Memora turns an AI execution into a receipt, checks that it still matches what happened, then gives teams a replay path for review and incident response.
Receipt
Capture the execution boundary around an AI call, agent step, or workflow.
RESULT + RECEIPT
Verify
Check that identity, input, output, and metadata still line up.
RECEIPT MATCHES
Replay
Reconstruct the receipt trail when a run fails or needs audit review.
INVESTIGATION-READY
Verify that the receipt still matches the run.
A receipt lets reviewers check whether the captured inputs, outputs, agent identity, and execution metadata still agree. Change the receipt after the fact and verification fails.
Edit any field to see verification fail. Simulated locally — run `memora replay verify` for the real check.
execution_id, agent, input, and output loaded
captured fields still match
ready for investigation
Receipt, verification, and replay stay connected without exposing infrastructure details in the developer workflow.
When something goes wrong, replay the receipt trail.
Investigate a failed, disputed, or suspicious run from the captured receipt trail. Review inputs, outputs, decisions, and state transitions without trusting the original runtime.
Replay Mode
View: System Flow
Receipts become an investigation trail.
When agents are active, the console shows recent receipt activity. Sample runs appear here until live data exists.
Sample receipt activity — live feed populates when agents are running.
Control which systems can issue receipts.
Memora gives each autonomous system an identity, credentials, and write policy so receipt generation is tied to the right agent instead of a shared log stream.
Registered system of record
Only this agent can write its receipts
Choose how strict verification should be
Add public auditability when you need it.
Teams can run Memora with internal receipt verification first, then enable external anchoring for deployments that need independent ordering or public audit evidence.
Receipt verification without a public network dependency.
Optional HCS ordering and registry commits for stronger external checks.
The backend interface is designed for additional anchoring targets.
Get your API key.
First 10,000 execution records free. Your first verifiable run in under five minutes.
Developer preview · schema may change between versions. Built in public · issues welcome on GitHub.