VerifiedState and peers like Engram give agents tamper-evident memory: signed assertions, hash chains, and MCP recall. DocImprint is a different layer — PDF/URL capture into signed evidence bundles with Merkle-cited answers, free offline verify, and optional Base L2 notarization.
Not a drop-in replace. DocImprint is not a replace for verifiable agent memory (VerifiedState, Engram). Those store beliefs and session knowledge; DocImprint proves what a document said at capture time.
Document ingest (PDF/URL) → evidence bundle with per-artifact SHA-256 Merkle citation proofs binding answers to source paragraphs Free offline verify + optional EAS notarize on Base — no vendor login required to check integrity
| Capability | VerifiedState | DocImprint |
|---|---|---|
| Primary object | Agent memory / assertions | Document evidence bundles |
| PDF / URL extract + OCR | Not the core job | POST /v1/extract |
| Citation proofs to source bytes | Partial (memory receipt) | Merkle verify-citation |
| Offline third-party verify | Yes (receipts) | Free GET /verify + ZIP |
| On-chain document notarize | Typically no | Optional Base EAS |
| MCP | Yes | Yes (20+ tools) |
Yes. Capture and prove documents with DocImprint, then store durable agent beliefs or session facts in a verifiable memory layer if you need long-lived grounded memory.
Engram is in the same category — provable agent memory with audit trails. Same distinction: memory store vs document capture evidence.
More comparisons: vs LlamaParse · vs Firecrawl · vs Unstructured · vs Lido · vs PEAC