Grounded. Keyless. Provable.
100%
Grounded by construction.
Before a word is written, your sources are mined for claims. A claim survives only if its quote appears verbatim in the source and every number in it appears in that quote. That verified ledger is the only substrate the writer is given.
0
API keys. No token budget.
Online, draft writes through the AI coding-agent CLI you already have — Claude, Codex, Copilot, Cursor, Grok and more — using that tool’s own logged-in session. Offline, it falls back to a local Ollama model and stays there.
C2PA
Provenance you can check.
Every article ships with a per-sentence attribution to the claims it rests on and a C2PA manifest binding it to its sources. draft --verify recomputes the digests and tells you whether a draft is exactly what the ledger was verified against.
Provenance and compliance — why shipping C2PA by default matters
How it works
Four phases, one grounded draft.
A research PDF in, a publication-ready Markdown set out — each phase reporting as it runs.
Read & section
A 62-page paper read and split in about 110 ms by a 10 MB binary. Poppler for PDFs, Docling for tables.
Extract & verify
One model call per section mines claims; each is checked against the source and dropped unless its quote is verbatim.
Write
One call arranges the verified ledger into an article in your house style. It arranges facts; it never sources them.
Attribute & sign
Per-sentence attribution and a C2PA manifest are written beside the set, so a reader can check the draft against its sources.
Every capability is a package
The claim gate, the engine seam and the pipeline are importable Go packages, tested without a network.
Publish-ready sets
Body, frontmatter and combined document, written side by side and regenerable without losing a curated field.
About draft
Why draft
A small local model will invent a plausible number. A cloud API will charge you for the privilege and want a network. draft takes neither risk.
Online, it writes through whichever AI coding-agent CLI you already have — Claude, Codex, Copilot, Cursor, Grok and more — using that tool's own logged-in session. No API key. No token budget. Offline, it falls back to a local model and stays there.
Either way, the model never gets to invent. Before a word is written, your sources are mined for claims, and a claim survives only if its quote appears verbatim in the source and every number in it appears in that quote. That verified ledger is the only factual substrate the writer is given. It arranges facts. It does not source them.
Point it at one paper or twenty. Each becomes its own draft, and every draft ships with a per-sentence attribution and a C2PA manifest so a reader can check it against its sources.
From paper to post. Grounded.
Point it at one paper or twenty. Each becomes its own draft, and every sentence can prove itself.
