Get product feedback from synthetic personas, inside Claude

Hey Indie Hackers,

Launching Cambium AI PM MCP today. It's a way to get product feedback from synthetic personas inside Claude.

The problem: PMs can't run a user interview every time they want to sanity-check a feature, and asking Claude what it thinks gives you Claude's opinion, not your customer's. Real customer research is slow, expensive, and scales badly. Most early-stage decisions get made on gut feel as a result.

What it does: Cambium AI builds synthetic personas from public demographic and market data. The MCP is the Claude integration that brings them into your sessions. You give it an idea, launch plan, or backlog item, and it runs the research through a generated panel of personas, mixes in web research and demographic context, and returns a report.

A few things it's useful for:

- Validating a new feature before committing eng time, by seeing how different segments react to the concept

- Pre-launch checks: running your positioning and messaging past the personas you're targeting and catching obvious misses

- Backlog triage: getting a sense of which items resonate with which segments when you have to cut scope

- Pre-mortems: asking the personas where they'd churn or stall

It also has a memory layer that carries business case, decisions, and project phase between sessions, so the personas aren't reacting in isolation each time.

It's an MCP server, so it works in Claude Desktop, Claude Code, or anywhere else that speaks MCP. Free during beta.

Obvious caveat: synthetic personas aren't real users. The value is in killing weak ideas fast and pressure-testing direction before you commit, not replacing customer conversations entirely.

More information: https://www.cambium.ai/product-manager-mcp