Data Refinery · Provenance · Evidence Reliability
Verifiable mineral
data
AssayChain sources, sanitizes, and delivers mineral data — assay reports, USGS commodity records, district history, and market context — as machine-readable, quality-graded JSON with cryptographically verifiable provenance. The core research is evidence reliability — measuring whether extracted mineral evidence can be trusted across five dimensions (fidelity, completeness, consistency, provenance, evidentiary sufficiency), with published error rates. Every data point independently verifiable.
Mission
Mineral data should be structured, open — and verifiable
Government and laboratory mineral data is technically public, but it's buried in PDFs, scattered across agency sites, and almost impossible to query programmatically. No API. No machine-readable format. No provenance chain linking raw report to delivered data.
Assay results, drill logs, and geological records are equally fragmented — scattered across lab certificates and internal reports. The next operator starts from scratch.
AssayChain exists to fix that: extract and structure the data, sanitize it against locked schemas, record a cryptographic provenance proof for each result, and serve it via a pay-per-call API that any AI agent or developer can access without accounts or subscriptions.
Public mineral and assay data is technically available, but it is locked in PDFs, agency portals, and lab certificates that resist programmatic access and independent verification. AssayChain applies open-infrastructure principles — cryptographically verifiable provenance, content-addressed storage, and access without a central gatekeeper — to mineral data.
AssayChain uses an open HTTP 402 micropayment protocol so software agents can purchase data programmatically, without accounts or subscriptions. A client pays per request with a signed authorization and receives the data — no OAuth flow, no API-key management. This makes verifiable data available to autonomous systems at the point of need.
What we deliver
The data refinery for mineral records
Four core surfaces, each with cryptographically verifiable provenance: extraction of assay documents into clean structured JSON, USGS critical-mineral benchmarks, historical mining district reports, and daily market context.
Assay extraction
The refinery's wedge. Raw assay certificates, drill logs, and geology PDFs become structured, quality-graded JSON (A–D) with full provenance (site, county, state, assayer, sample dates). Live-verified on 1937 DOGAMI fire-assay sheets (Au 0.64 oz/t) and USGS Circular 592 (57 samples).
USGS commodity benchmarks
Official USGS Mineral Commodity Summaries for 20 critical minerals — production, reserves, top producers, China share — served as machine-readable API responses with verifiable provenance, so AI agents can query USGS data programmatically instead of parsing PDFs.
Historical districts
Curated US/MX mining district reports and knowledge-graph relationships — deposit geology, historical production, MRDS references. 182+ districts across 11 US states and 2 Mexican states, plus commodity relationship data.
Market context & derived scores
Daily price snapshots (spot, BDI, WTI, FX) plus assistive AI-derived scores — risk, resilience, ESG, compliance, ore-potential. The derived layer is labeled assistive research context, never a substitute for professional review.
Provenance
Every record cryptographically verifiable
The provenance pipeline is the core: source data is sanitized against locked schemas, stored in content-addressed storage, and recorded with a cryptographically verifiable attestation. Each paid response carries an attestation_uid and ipfs_cid — a permanent, tamper-evident link from what you receive back to the original source record.
Pipeline proof is real and live-verified end-to-end: the pipeline has structured and attested real public documents — 57 USGS fire-assay samples graded A, and 1937 Oregon DOGAMI fire-assay sheets recovered from a degraded scan — proving the schema, attestation UIDs, and content-addressed provenance are genuine rather than vaporware. The commercial product is extraction of assay and geology documents, plus the attested ground-truth core.
A public, low-cost attestation registry keeps verification inexpensive at production scale. A standard schema registry creates typed, cryptographically verifiable records — no custom contract required. Both are built for production-scale usage, not experiments.
Storing full records on every write would be prohibitively expensive. Instead, each record is stored in content-addressed storage — a file address is derived from its content, so any change produces a different address. A compact cryptographic attestation holds a permanent pointer to it. Anyone can retrieve the original data and verify it matches what the API returns.
Who we are
Quetzal Collective LLC
AssayChain is a product of Quetzal Collective LLC — a Missoula, Montana company building verifiable data infrastructure for the mineral data space. We source data from government agencies (USGS), public filings, and third-party research; sanitize it against locked schemas; extract and structure the rest; and deliver it via a verifiable, pay-per-call API.
Team: a four-person group — founder-PI (software engineering; agentic integration: MCP, machine-to-machine payment, AI orchestration), working with a PhD geologist for domain validation of extraction outputs, deposit contexts, and compliance-language accuracy. The geologist's domain knowledge is what keeps the geological content defensible.
Advisory build-out (in progress): adding a cryptography/distributed-systems advisor, a mining-finance / Qualified Person advisor, and a fractional commercial lead, and pursuing STTR eligibility with Montana Tech as research partner.
Data streams are live: USGS Mineral Commodity Summaries for 20 critical minerals, assay extraction from PDFs and images, curated US/MX historical mining district reports, daily market context bundles, and company critical-minerals exposure profiles assembled by a multi-agent swarm. Intelligence endpoints (risk, compliance, knowledge-graph, procurement, deposit-model, ore-potential, enrich) synthesize attested data into structured outputs.
Everything is verifiable: provenance schema UIDs are public, source data is content-addressed and openly retrievable, and the payment gate is mechanical — pay and receive data. No gatekeeping.
Early revenue (~USD 500/month) with no analyst desk or enterprise sales team yet. Compliance and risk outputs are assistive research context — always require Qualified Person / Competent Person review before any regulatory use. We state this plainly because trust starts with honesty.
The provenance pipeline is demonstrated on real public data — USGS Circular 592 (57 samples, grade A) and 1937 DOGAMI fire-assay sheets. That data is not marketed as a product; it proves the pipeline works end to end. The commercial products are extraction and the attested ground-truth core.
Extraction is validated against real ground truth: 57 USGS fire-assay samples graded A, and 1937 Oregon DOGAMI fire-assay sheets (Au 0.64 oz/t) recovered from a degraded scan. We are building a published, Qualified Person–validated benchmark with field-level error rates and confidence intervals — the measure the product is graded against.
Designed for AI
For AI agents and LLMs
AssayChain ships with Model Context Protocol (MCP) server registration on Anthropic, Glama, and Smithery. Register the MCP endpoint to give any Claude, ChatGPT, or Groq agent access to the data ecosystem.
MCP Tools
Free: corpus.search, extract.estimate, sales.ask, benchmark.sample, benchmark.commodity
Preview free / paid via REST: criticality.crosscheck, china_trade_control, supply_share, stockpile_level, supply_resilience, esg_profile, blm_claims_data, geophysics_surveys, active_operations, earth_mri_focus_areas
Paid ($0.10–$0.50): district.history, extract.run
No API Keys
Agents pay directly through an open micropayment protocol (USDC) or by credit card. No accounts, no subscriptions, no rate-limit keys to manage. Your agent can call 1 endpoint or 1000; billing is per-call, not per-seat.
Verifiable Receipts
Every paid call returns a cryptographically signed proof of payment that can be independently verified. Forward receipts to your auditor. Useful for compliance and audit trails.
Commodity + Compliance Use Cases
Query 20 critical minerals for ESG disclosure, UFLPA sourcing evidence, EU Battery Reg 2023/1542 compliance, CBAM tariff impact, DoD domestic mineral assessments, or M&A due diligence. Agent-friendly schemas with clear examples.
Model disclosure
Extraction runs on Gemini 2.5 Flash; the enrichment/sanitization pass on DeepSeek or GLM; Q&A synthesis on Groq. Every model in the pipeline is disclosed on the Security page. We are migrating compliance-sensitive enrichment to US-hosted or self-hosted inference.
Persistent Cache — $0.10 Reuse
Extract results and enrichment profiles are now saved to Firestore + content-addressed storage. Subsequent calls for the same source URL or company within 7 days return cached results at $0.10 — 80% cheaper than fresh compute.