> ## Documentation Index
> Fetch the complete documentation index at: https://productos.dev/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Claude

> Connect Claude to your ProductOS project over MCP so it plans, specs, and pressure-tests decisions against your real product context.

The ProductOS Claude integration connects Claude to your project over [MCP](/integrations/mcp), so Claude reasons over the real product context: no copy-paste, no re-explaining the product every session, no drift.

## What Claude sees

* **The PRD.** Claude reads the live requirements: scope, user stories, and acceptance criteria, so its plans and answers are grounded in what you are actually building.
* **The research.** Market and user findings are part of the context, so Claude reasons with the evidence behind the product instead of generic assumptions.
* **The design.** Flows and UI decisions are visible to Claude, keeping its suggestions consistent with the intended experience.

## Why connect Claude

Teams reach for Claude to plan, spec, and reason through hard product decisions. The missing piece has always been context:

* **No copy-paste.** No more pasting the PRD into a chat and re-explaining the product every session. The MCP connection lets Claude read the live context straight from ProductOS.
* **Grounded reasoning.** When Claude can see the requirements and the research, its output stops drifting. You get analysis and drafts that hold up against the actual product definition.
* **Results flow back.** MCP is bidirectional, so updates Claude makes are written back into the shared context where the rest of the pipeline can use them.

## Connect in three steps

<Steps>
  <Step title="Open your ProductOS project">
    Choose the project you want Claude to reason over and open its integration settings to reveal the MCP connection.
  </Step>

  <Step title="Add the MCP server to Claude">
    Connect the ProductOS MCP endpoint in Claude's connector settings and authorize it. Claude can now read your project as a live context source.
  </Step>

  <Step title="Reason over live context">
    Ask Claude to refine the spec, pressure-test a flow, or draft the next iteration.
  </Step>
</Steps>

## What connected Claude is good at

* **Refine the spec in place.** Ask Claude to sharpen user stories or acceptance criteria and it edits against the real PRD, not a copy.
* **Pressure-test decisions.** Have Claude challenge a product choice using the research that is already in your project context.
* **Plan the next iteration.** Draft the roadmap for what comes next with Claude reasoning over where the product actually stands today.

## Claude vs Cursor: which one?

Both integrations expose the same project context over the same protocol. The difference is where you work: Cursor is the editor, where the connected context steers code generation; Claude is where you plan, spec, and reason before and between builds. Many teams use both, and because they read one shared context, nothing diverges between them. See the [Cursor integration](/integrations/cursor).

<Note>
  Whatever you produce with Claude connected remains yours. The code side of the pipeline syncs to your own repository through the [GitHub integration](/integrations/github).
</Note>