GOVERNED RUNS
PUBLIC VERIFICATION
Signed run receipts for AI agents.
When an AI agent makes a decision, a run receipt records what it actually did across the whole task: which rules it consulted, which version of each, sealed into one signed receipt that anyone can verify.
It is the evidence layer for autonomous workflows. Built for support bots, agent pipelines and governance teams who need to show their work later.
START FREE TRIAL →
OPEN THE TOOL →
How a run receipt works
Four steps, from the moment a task starts to a review-ready evidence pack.
01
Open a run
When your agent or workflow starts a task, it opens a run. The run is a container for everything the agent consults and decides from that point until it finishes.
02
Record each turn
As the agent works, it records one entry per turn. Fetching a rule and recording a verified entry can happen in a single call, and each verified entry pins that rule's exact version and integrity fingerprint at that moment. An entry can also note the model used and a private hash of the end user.
03
Seal it into one receipt
When the task is done, the agent seals the run. Sealing produces one signed run receipt that fixes the whole sequence: every rule consulted, in order, at the version it held when the agent relied on it.
04
Export a governance audit pack
From a sealed run you can export a governance audit pack: the receipt, every entry, an independent tamper check, and a map of the record-keeping obligations the run provides evidence toward. Hand it to a reviewer as-is.
What the receipt proves
A receipt is only useful if a third party can trust it without trusting us. This is how.
🧾
WHAT WAS CONSULTED
Every entry pins the rule version and integrity fingerprint the agent actually relied on, so the receipt reflects the law as it stood at the moment of the decision, not as it reads today.
🔓
INDEPENDENT TO VERIFY
A sealed receipt verifies against our published signing key with no Bidda account. A reviewer, an auditor, or a counterparty can confirm it independently.
🌲
APPEND-ONLY HISTORY
Every signed record is added to a public, append-only transparency log built the same way as Certificate Transparency (RFC 6962). Anyone can prove a receipt is in the log and that the log was never re-ordered or back-dated.
The obligations it helps you evidence
Record-keeping is written into modern AI regulation. A run receipt gives you something concrete to put against it.
EU AI ACT, ARTICLE 12
Logging and record-keeping
Automatically generated records of the events across an AI system's operation.
EU AI ACT, ARTICLE 26
Deployer obligations
Deployers keeping the logs their high-risk AI systems generate.
ISO/IEC 42001
AI management systems
Documented evidence that AI operations were carried out under a governed process.
NIST AI RMF
Govern and Measure
Traceable records that support the govern, map, measure and manage functions.
A run receipt is evidence you can put in front of a reviewer. It is a record of what an agent consulted, not a determination that an obligation is met.
Export it where your reviewers work
A sealed run is readable by id and exports to the formats GRC and audit tooling already understand.
READ A SEALED RUN
GET /api/v1/runs/{run_id}
Public by id. Re-check the receipt against our signing key with no account.
GOVERNANCE AUDIT PACK
get_audit_pack
The receipt, every entry, an independent tamper check, and the obligation map, over the API or MCP.
NIST OSCAL EXPORT
oscal_assessment_results
Export a governed run as a NIST OSCAL assessment-results document and drop it straight into an OSCAL-aware pipeline.
COMMON QUESTIONS
What is a run receipt?
A run receipt is a single signed record that covers a whole conversation or task rather than one decision. An agent opens a run, records one entry per turn as it consults rules and answers, then seals the run into one receipt. The receipt is a time-stamped, tamper-evident summary of what the agent relied on across the task.
How is a run receipt different from a single signed record?
A signed record captures one decision and the rules behind it. A run receipt captures an entire run: many turns, many consulted rules, in order, sealed together. Use a signed record for a single decision and a run receipt when you need evidence across a full support conversation, agent task, or pipeline execution.
Can someone verify a run receipt without a Bidda account?
Yes. A sealed run is readable by its id and verifies against our published signing key with no account and no payment. Every signed record is also added to a public append-only transparency log, so a reviewer can confirm the receipt is included and that the log was never altered.
What is in the governance audit pack?
The audit pack bundles the sealed receipt, every recorded entry, an independent tamper check, and a map of the record-keeping obligations the run provides evidence toward, such as EU AI Act Article 12 logging and Article 26 deployer duties, ISO/IEC 42001, and the NIST AI RMF. It is built to drop straight into a review file.
Can I export a run as OSCAL?
Yes. A governed run can be exported as a NIST OSCAL assessment-results document, the machine-readable evidence format that GRC and audit tooling consumes. Reviewed controls list the obligation nodes consulted, observations pin each to its version and integrity hash, and the export carries an independent integrity self-check.
Is a run receipt a statement that I am compliant?
No. A run receipt is evidence of what an agent consulted and relied on, which you can put in front of a reviewer. It is a record, not a determination of compliance. A qualified professional still reviews the underlying obligations and your controls.
Give your agents a record they can prove.
Run receipts are included with every subscription and a free trial counts. Wire them up over the API or MCP, or try the tool in your browser.
START FREE TRIAL →
DEVELOPER DOCS →