If an agent did it, there's a receipt.

This actor runs an AI agent task and returns the result plus a cryptographically signed receipt of every action the agent took: each model call and tool call, in order, signed so you can verify it offline.

• A completed task. Give it a task and your own LLM API key (Anthropic or OpenAI); it runs an agent loop until done or it hits a step limit.
• A signed receipt. The full action log, plus the Apify run's identifiers, is signed with a fresh Ed25519 key. Check the signature offline with nothing but the receipt, and confirm nothing changed after signing.
• A platform-attributable record. The receipt is also written to the run's own dataset and key-value store, which only the running actor can write to, so provenance can be confirmed against the Apify API.
• Free verification. Re-run in "verify-receipt" mode to check any receipt's signature. This mode is free.

What it proves: verification confirms the action log was not altered after signing. It does not by itself prove the run happened on Apify, since anyone can sign a fabricated log; provenance means checking the receipt's actorRunId against that run's record on Apify. No SOC 2, ISO, or other compliance claim.

• mode: run-task (default) or verify-receipt
• task: the agent's task prompt (run-task)
• llmProvider: anthropic or openai (run-task)
• llmApiKey: your own key, never stored or logged (run-task)
• model, toolAllowlist, maxSteps: optional
• receipt: a prior receipt to check (verify-receipt)

• Governed action: $0.02 (per model/tool call recorded)
• Governed run completed: $0.08
• Receipt verification: free

The price covers orchestration and the receipt, not model tokens, so you keep full control of your model spend.

Built by RevealUI Studio.