What it does
Context Factory is a chat-based content engine built around one idea: the model should know your business before it writes for it. As you work in the chat, the assistant identifies business facts in the conversation (a pricing decision, a customer segment, a competitor's move, a positioning statement) and proposes them as knowledge items. A second review step, the Guard Agent, checks each proposal against what is already saved and flags contradictions or duplicates before anything is accepted.
Once accepted, that knowledge is context for every future session. Ask for a landing page draft, a comparison table, or a diagram, and the assistant writes from your facts rather than from generic training data.
Other capabilities:
- Multiple model providers. Bring your own key for the provider you prefer, or use the default.
- Artifacts. Save documents, code snippets, and Mermaid diagrams from any conversation and find them later.
- Live HTML preview. Prototypes the assistant generates render in a sandboxed preview.
- Cost visibility. Each message shows what it cost, with a full transaction history.
Who it's for
- Content marketers who spend more time correcting a first draft than they would have spent writing it, because the draft did not know the product.
- Product marketers maintaining positioning across many pieces and needing every one to agree with the canonical version.
- Small GTM teams where one person writes for sales, marketing, and the website and needs a single source of facts.
The pattern it replaces is the pasted-in brief. Instead of re-explaining your company at the top of every prompt, you explain it once, review what the assistant understood, and let it accumulate.
How it works
- Start a session. Open a factory session and describe what you are working on.
- Work in the chat. As you discuss the product, market, or customer, the assistant proposes knowledge items it has noticed.
- Review proposals. The Guard Agent reports whether each item conflicts with existing knowledge. You accept, edit, or reject.
- Generate. Ask for the content you need. Drafts draw on the accepted knowledge base and cite the items they used.
- Save artifacts. Keep the outputs worth keeping. They stay attached to the session and the knowledge that produced them.
The app is in the queue for public launch on its subdomain. Join the waitlist on the launcher to be notified when it opens.
Related resources
- Guide: How to create content from sales calls — where the facts that ground good content come from.
- Template: Messaging framework — the structure to load in first.
- Glossary: Voice of customer