Section-level regeneration
Don't like the intro? Regenerate just that section. Proof block weak? Regenerate. The rest stays untouched. No more burning a full draft to fix one paragraph.
PLATFORM
Source-grounded articles in your voice.
The draft engine produces publishable Markdown articles from the brief — section by section, citation-backed, with brand tone respected and forbidden phrases blocked.
Built for content teams that ship
Real claim grounding
Each generation pass computes honest grounded/flagged/blocked claim counts from the actual content. No hardcoded "12 sources" badges — the numbers reflect real citations and risky superlatives.
Forbidden phrases enforced
AI clichés (delve, leverage, tapestry, ever-evolving) and your own banned list are blocked at the quality gate. Output ships clean.
Brand voice from examples
Paste 1-3 paragraphs of your best writing in Settings → Brand. The draft engine uses them directly as voice anchors.
UNDER THE HOOD
Drafts route through OpenRouter with Anthropic Claude Sonnet 4.6 as the default writer. Temperature is held low, max tokens are sized to the brief's target word count, and a quality gate runs on every output — minimum length, real section count, no planning notes, no workspace label leak, no forbidden phrases.
Start writing with the loop.
Pro plan. 50 drafts a month. Cancel any time.