Document infrastructure for the AI era

Your LLM can finally read tables, math, and data.

tdoc turns any PDF, DOCX, or HTML into a .tdoc — a typed, queryable, signable document file. Every number carries a declared type. Every file carries an Ed25519 signature. Every archive is byte-deterministic — same input, same bytes out, always. Ship in a single REST call.

Start free Read the API Self-host
< 2s median latency, 10-page PDF
100 docs/mo free, forever
Ed25519 signatures, zero wall-clock drift

What you get

Typed every node

Cells declare data-type="pvalue" / measure / currency. Your downstream code never guesses.

Query with AQL

"Find every p-value < 0.05 in this 300-page paper" — one line, no LLM needed.

Signed & verifiable

Ed25519 over canonical bytes. Tamper once, verification fails forever.

Byte-deterministic

Same input → same output, guaranteed. Hash it, sign it, audit it.

Accessibility built in

alt-text coverage, equation descriptions, reading order — not an afterthought.

Open format

Run the SDK offline. Your data never needs to stay on our servers.

Two lines to structured data

curl -X POST https://api.tdoc.xyz/v1/structure \
  -H "Authorization: Bearer $TDOC_KEY" \
  -F "file=@paper.pdf" \
  -F "title=My Paper"

Returns an AXON document tree. Run /v1/query for AQL, /v1/sign for Ed25519, /v1/verify to validate a third-party archive. Full OpenAPI spec at /docs.

Pricing

Start free. No credit card. Upgrade when your project does.

Free

$0/mo
  • 100 documents / month
  • All parsing features
  • AQL queries included
  • Community support
Get API key

Pro

$29/mo
  • 2,000 documents / month
  • Email support
  • No hard overage — upgrade anytime
  • Signing (Team+)
Start Pro

Team

$99/mo
  • 10K documents / month
  • Ed25519 signing + verification
  • Priority email (24h SLA)
  • Overage: $0.02 / doc
Start Team

Scale

$499/mo
  • 100K documents / month
  • Self-hosted option
  • Slack support channel
  • Overage: $0.01 / doc
Talk to us

Why now

Every AI team is losing 20–40% of document quality to broken PDF ingestion. Tables become strings. Equations become gibberish. P-values lose their type. tdoc gives LLMs the same structured view a database has — so your agents can cite a number, not just see it.