We turn scattered product evidence (lab reports, certifications, test methods, manufacturer specs) into structured, source-linked records that AI agents and commerce platforms can query.
The missing layer
Catalogs tell machines what a product is. TrueChoice tells them what evidence exists behind it, and what doesn't.
How it works
The evidence lifecycle
We build and maintain the evidence. The judgement about what to recommend stays with the agent and the platform that runs it.
Certificates of analysis, lab reports, certifications, manufacturer specs, labels
Pulls each document and records who issued it, when, for which lot, and by what method.
Which sources your platform trusts
Label claims, listings, spec sheets
Splits each product into claims, measurements, analytes and methods, each linked back to the page it came from.
How claims are presented to shoppers
Test results, certification scopes
Checks whether each claim has a supporting measurement and records the match or the gap.
Thresholds, and what counts as enough
The absence of a document
Records absent evidence as absent, so a missing heavy-metals panel never reads as a clean one.
How much a gap should weigh
The structured record
Returns evidence, coverage, provenance and dates in one object the agent can reason over.
The recommendation and the words to the shopper
New lots, renewed and expired certifications
Re-dates evidence as it changes, so an agent can see that a result is 81 days old.
When old evidence stops counting
The record
Same product, same shelf. One record says it's ready to sell. The other says what has been shown, how, and what nobody has checked yet.
Example record for illustration
For agent builders
TrueChoice runs as an MCP server. Your agent calls it the way it calls any other tool and gets structured evidence back.
get_product_recordThe full record for a product: claims, results, sourcesget_evidenceThe documents and measurements behind one claimget_coverageWhat has been tested, what hasn't, and how recently// agent → TrueChoice { "product": "MGG-400-120CT" } // TrueChoice → agent { "coverage": { "found": 3, "of": 6 }, "signals": [ { "id": "heavy_metals", "status": "tested", "source": "CoA #EU-48221", "date": "2026-08-14" }, { "id": "micro_panel", "status": "absent" }, { "id": "stability", "status": "absent" } ], "missing_evidence": "flagged_explicitly" }
One layer, many decision surfaces
Ask for the evidence behind a product before recommending it, and say plainly when evidence is missing.
Put verified product intelligence behind search, comparison and checkout without building a lab-data team.
Get credit for the testing you already paid for, in a form agents can read and cite.
FAQs
Product-intelligence infrastructure. We collect product evidence, structure it into records with sources and dates, and make those records queryable by AI agents and commerce platforms.
No. TrueChoice returns the evidence and where it came from. The agent or platform decides how to weigh it for the person it's helping.
Certificates of analysis, independent lab reports, certification bodies, and manufacturer specifications. Every value in a record links to the document it came from.
It's recorded as missing. An untested contaminant panel shows up as "not tested", so an agent never mistakes silence for a clean result.
Brands can fund the intelligence for their products. Platforms license access to query it. Shoppers never pay.
We're onboarding design partners now. Email us with what you're building and we'll set up a working session.