docimprint

Merkle proof explainer for document citations

See how binary Merkle trees let agents prove a specific paragraph came from a signed evidence bundle — O(log n) proof size.

How it works

Each document chunk is hashed (SHA-256) and placed as a leaf. Parent nodes hash pairs of children until a single Merkle root commits to all chunks. The root lives in manifest.json and is covered by the platform signature.

  1. Extract document → N chunks with chunk_id
  2. Build Merkle tree → store merkle_root in manifest
  3. Agent cites chunk_id → API returns sibling hashes (proof)
  4. Verifier recomputes root offline — no trust in DocImprint

Chunks

16

Proof hashes (≈)

4

Proof size (bytes)

~128

Need API access? Get a free API key →

Related