
Engines · Text and writing
The engines that write every word. In your voice. Tied to real facts.
9 text engines write posts, pages, emails, ads and review replies, score leads and transcribe speech. Claude on an owned seat goes first at $0 per job; local models are the floor.
demo businessWhat they do
More than writing posts
Most OS output starts as words: captions, blogs, service pages, emails, Google ads, review replies, scripts, texts. Text engines write it; Writers' Room rules check it before you see it.
Behind the scenes, a local model scores every lead from 0 to 100, hottest first. Whisper transcribes captions and calls. bge-m3 makes your knowledge base searchable, so your assistant cites sources.
Claude, on an owned seat, writes first: drafts, long-form, prompts, brand-voice rewrites, $0 per job. Overflow goes to a backup writer, pay as you go, never silent default. Ollama with Qwen3 8B drafts offline: the floor.
The catalog
All 9 text engines
Each one with its lane. Tap any engine to see what it's used for.
- ClaudewritingOwned seat
- Claude APIwritingMetered · asks first
- a backup writerwritingMetered · asks first
- Qwen3 8BwritingFree engine · $0
- Llama 3.1 8BwritingFree engine · $0
- Qwen 2.5 72BwritingFree engine · $0
- Llama 3.1 70BwritingFree engine · $0
- bge-m3writingFree engine · $0
8 in this group, from the Engine Room manifest. Lanes: free engines first ($0), owned seats, metered engines that ask first. Product fact
Who writes what
The job each engine is picked for
- Claude (writing seat)Primary writer: drafts, long-form, prompts, brand-voice rewrites. Our client-safe voice winner.Owned seat
- Ollama · Qwen3 8BOffline drafting. The always-available floor.Local · $0
- Ollama · Llama 3.1 8BLead scoring from 0 to 100, plus quick local reasoningLocal · $0
- Qwen 2.5 72B (planned)Long-form writing on our free enginesFree engine · $0
- Llama 3.1 70B (planned)Heavier reasoning on our free enginesFree engine · $0
- bge-m3Makes your knowledge base searchable (embeddings)Local · $0
- Whisper · faster-whisperSpeech to text: captions and call transcriptsLocal · $0
- OpenRouter · SonnetFall-through when the writing seat runs out. Never the silent default.Metered · asks first
- Claude APIHigh-volume batches beyond the seatMetered · asks first
Free engines first. Premium engines only when you approve, with the cost shown before anything runs.
The rules
The rules every engine writes under
The engine drafts. The OS holds it to the same rules, whichever engine wrote it.
| Rule | What it means for you |
|---|---|
| Claims binding | Any number or promise has to be tied to a real fact on file, or it's blocked or softened. |
| Your voice | Written from your Brand DNA, with a voice match against your own copy and tone presets like Blunt, Numbers-first, No jargon and Warm neighbor. |
| Anti-slop score | One-tap edits: shorten, expand, change the tone, more human. |
| Platform limits | The first 125 characters show before “more”, so the hook goes first. X posts thread themselves past 280 characters. |
| Search ready | Every piece gets a search title and description (50-160 characters) and alt text for its images. |
| Customer data guard | Drafts with personal info route to a private lane and are locked off the free tier. |
| Texts and email | Text messages are written TCPA-safe (the US texting law), and the checks look for an opt-out and a CAN-SPAM footer (the US email law). |
| Last check | Nothing goes out until the Screening Room passes it and you approve it. |
Product fact: rules from the Writers' Room and Screening Room.
Inside the OS
Where the text engines work
Blogs, emails, ads, captions, scripts, texts, landing copy and review replies, with a creativity dial from 1 (safe) to 11 (wild).
Articles and service pages written to be found in search and in AI answers.
Whisper times every word for karaoke captions and SRT/VTT files. Plan the cut from the transcript; exporting your edited video is on the way.
Recency ranks the callback queue.
faster-whisper hears the caller so the line routes by what they say. Emergencies ring through to your phone. Live once your phone line is connected.
Your voice, offer and facts: what every engine writes from.
Cost and license gates. The writing seat and local models cost $0 per job. Metered text (a backup writer, Claude API) runs only as fall-through or big batches, asking first. 1 rule: quoted on the sum, charged on what returned, released on cached and skipped work. Non-commercial engines are blocked.
Questions owners ask
Will it sound like a robot?
It writes from your Brand DNA and checks against your own copy. An anti-slop score flags generic lines, “more human” is one tap away, and you approve every piece.
Can the AI make up facts about my business?
Not past the claims check. Any number or promise must tie to a fact on file, or it's blocked or softened before you see it.
Does my customers' info go into a public AI?
No. Drafts with personal info route to a private lane and are locked off the free tier.
What happens if the writing seat is busy?
The job falls to the next engine. Local models are always there at $0. Paid fall-through runs on your terms, never the silent default.
Keep going
Related pages
Try the OS
Our free trial is starting soon.
The same tools we use for our done-for-you clients. Our 7-day free trial is starting soon. Until then, book a call and a real person walks you through the OS. $997 a month, first month $497. No contracts.

YOUR TEAM runs it with the OS, or OUR TEAM runs it for you with Done-For-You. Not sure? Which one fits me?





