docimprint

Frequently asked questions

Common questions about verifiable document memory, evidence bundles, x402 payments, and agent integration.

What is verifiable document memory?

Verifiable document memory is a persistent, cryptographically auditable layer where AI agents store document captures — not just embeddings or summaries, but tamper-evident evidence bundles with manifest SHA-256 hashes, artifact digests, optional Merkle citation proofs, and chain-of-custody metadata. Unlike RAG, every claim an agent makes can be traced back to an immutable, verifiable source.

How is DocImprint different from Firecrawl or LlamaParse?

Firecrawl returns Markdown chunks. LlamaParse returns structured JSON with optional citations. DocImprint returns a cryptographically verifiable evidence bundle — bundle_id, manifest SHA-256, EIP-191 signature, and per-artifact hashes — plus a free offline verify endpoint. No other tool lets you prove, after the fact, that a document has not been modified since the agent captured it.

Why not just use RAG?

RAG stores embeddings, not evidence. An embedding can be regenerated from any version of a document — there is no proof that the text was captured at a specific time and has not been modified. DocImprint evidence bundles are cryptographically bound to the original document and timestamped with an optional on-chain notarization. For high-stakes domains (legal, finance, compliance), the difference between "the model said so" and "here is the tamper-evident source" matters.

What is an evidence bundle?

An evidence bundle is the output of POST /v1/extract. It has a unique bundle_id (ev_…), a manifest listing every artifact (Markdown, screenshot, PDF, OCR text, structured data) with its SHA-256 hash, and an EIP-191 signature over the manifest. Artifacts are stored in DocImprint-managed R2. Anyone with the bundle_id and owner credentials can verify for free with GET /v1/extract/{id}/verify.

How do I verify a bundle without calling the API?

Download the bundle ZIP with GET /v1/extract/{id}/download. The ZIP contains the manifest.json, all artifacts, and the manifest signature. Recompute SHA-256 on each artifact, compare to manifest, then verify the EIP-191 signature against the published secp256k1 key at GET /v1/keys. No network call to DocImprint required — the verification is fully offline.

What is the x402 payment protocol?

x402 is an HTTP 402 Payment Required flow for micropayments. The client sends a request; the server responds 402 with payment requirements (amount, asset, recipient, network). The client signs a USDC EIP-712 permit transfer and retries with an X-Payment header. The server verifies and processes in a single round-trip. No account, no API key, no subscription required — a wallet address is the identity.

Do I need a wallet to use the API?

No. You can use API key authentication instead — purchase a plan at docimprint.com/pricing and pass Authorization: Bearer dr_live_… with every request. API keys come with monthly credit allowances (100 credits free, 20,000 credits on Pro). x402 wallet payments are the alternative for agents that prefer pay-per-call USDC on Base.

What's the difference between lean and bundle extract?

Bundle extract (the default) stores all artifacts — screenshot, PDF, Markdown, OCR, structured data — in an evidence bundle with cryptographic hashes. Lean extract (?bundle=false) returns only the AI analysis text and skips artifact storage. Lean is cheaper and faster; bundle is required when you need offline verification, legal hold, or downstream citation proofs.

How does claim-check mode work?

POST /v1/extract with mode=claim-check and a claims array. DocImprint fetches and extracts the document, then runs each claim against the text, returning status (supported | contradicted | not_found), a verbatim evidence quote, paragraph references, and a confidence level. Claim-check works on any mode (lean or bundle) and scales with claim count.

What happens if a bundle is tampered with?

GET /v1/extract/{id}/verify recomputes the manifest SHA-256 and all artifact hashes from storage and compares them to the signed manifest. If any byte has changed, the endpoint returns 409 Conflict. The EIP-191 signature also fails independently. Both checks must pass for a bundle to be considered valid.

When should I use legal hold?

PUT /v1/extract/{id}/hold blocks the DocImprint retention garbage collector from ever deleting that bundle. Use it when a document is relevant to active or anticipated litigation, a regulatory audit, or any scenario where deletion would be sanctionable. Legal hold is free and can be set immediately after capture.

How does notarization work?

POST /v1/extract/{id}/notarize anchors the manifest SHA-256 hash on Base L2 (Ethereum Layer 2). The on-chain transaction hash is returned and stored with the bundle. Anyone can verify the hash on-chain without trusting DocImprint. Notarization costs $0.05 per bundle and is independent of capture — you can notarize any existing bundle.

What is MCP and how do I use DocImprint with it?

Model Context Protocol (MCP) connects AI hosts to DocImprint at https://api.docimprint.com/mcp. An API key is required (Authorization: Bearer dr_live_...). Claude Desktop cannot use a remote URL in claude_desktop_config.json — use the mcp-remote stdio bridge (see /docs/agents#mcp). Cursor may use url + headers. Tools include extract_text, extract_tables, parse_invoice, check_claims, extract_structured, verify_bundle.

What extract modes are available and what do they cost?

Bundle (default, $0.075): full artifact capture with cryptographic hashes. Lean ($0.010): text only. Summarize ($0.018): 3-sentence summary + key points with citations. Q&A ($0.022): answer a specific question with paragraph citations. Claim-check ($0.025): verify a list of claims against the document. Translate ($0.040): full document translation. Structured ($0.025): extract fields from a schema. All modes support both PDF uploads and hosted URLs.

What is BYOS (bring your own schema)?

BYOS means extract-structured mode: you pass a JSON schema and DocImprint returns fields in that shape with citations. It is not customer-owned S3 storage — all artifacts are stored in DocImprint-managed R2 unless you download the bundle ZIP.

I lost my API key. Can I retrieve it again?

No. API keys are shown exactly once at checkout. There is no self-serve rotation. If you lost a Free key, sign up with a new email or contact info@sawftware.co. If you lost a Pro key, contact support or start a new Pro subscription at docimprint.com/pricing.

DocImprint vs RAG vs web scrapers

CapabilityTypical RAGWeb scraperDocImprint
Proof of unmodified captureNoNoSHA-256 manifest + EIP-191 signature
Offline verificationNoNoFree GET /verify + ZIP download
Merkle citation proofsNoNoO(log n) proofs (~320 bytes)
Agent pay-per-callN/AAPI key onlyx402 USDC from $0.010/call

Related