Connect Outcomify to Claude with the MCP server

Outcomify ships a built-in Model Context Protocol (MCP) server. Once connected, an MCP-aware assistant like Claude or Cursor can read your Opportunity Solution Trees and help you work on them in plain language — "what changed in the activation tree this week?", "find every untested assumption", "draft three opportunities under the onboarding outcome" — without you leaving the chat.

Because it speaks MCP, the same connection works across Claude Code, Claude Desktop, Cursor, and any other MCP client.

What your assistant can do once it's connected

The server exposes Outcomify as a set of tools, grouped by what they touch:

Read your tree

• list_trees, get_tree, get_node — load the structure and the details.
• find_nodes — search across opportunities, solutions, and assumptions (including by tag).
• what_changed and tree_history — see recent activity and how the tree evolved.

Draft changes for review

• propose_opportunity and propose_edit — suggest additions and edits as drafts that your team reviews and approves before anything lands in the live tree.

Edit directly (only if you grant write access)

• create_node, update_node, move_node, set_node_tags, add_evidence, set_outcome_metrics, and more.

You decide which of these are allowed — see permissions below.

Connect it in two minutes

The server lives at:

https://api.outcomify.com/mcp

It's protected by OAuth, so you never paste an API key. The first time your client connects, Outcomify opens a browser window, you sign in, and you approve exactly which scopes to grant. Client registration happens automatically (Dynamic Client Registration), so there's nothing to set up in the Outcomify dashboard first.

Claude Code

claude mcp add --transport http outcomify https://api.outcomify.com/mcp

Run /mcp inside Claude Code to finish signing in, then start asking about your trees.

Claude Desktop

Open Settings → Connectors → Add custom connector, name it "Outcomify", and paste the URL above. Approve the connection in the browser window that opens.

Cursor

Add it to your mcp.json:

{
  "mcpServers": {
    "outcomify": {
      "url": "https://api.outcomify.com/mcp"
    }
  }
}

Permissions: read-only by default

During sign-in you choose what to grant:

• outcomify.read — the assistant can see your trees but cannot change anything.
• outcomify.write — the assistant can create and edit nodes, and submit drafts.

If you only grant read access, the write tools simply aren't available. A good way to start is read-only: let the assistant explore and propose drafts, then review them in Outcomify and approve what's useful. Access tokens are scoped to your organization and to the Outcomify resource only, so a leaked token can't be replayed against other services.

A few prompts to try

• "Summarize the current state of our activation outcome tree and flag the riskiest untested assumptions."
• "Find every opportunity tagged Bug across all trees and list them by tree."
• "Based on these three interview notes, draft candidate opportunities under the onboarding outcome." (then review the drafts in Outcomify)
• "What changed in the retention tree in the last week, and who made the changes?"
• "Here's a screenshot of our Miro tree — recreate it in Outcomify, keeping the hierarchy." (More in import your Miro tree with a screenshot.)

Leading more than one team?

The same connection powers one of the most useful things a product leader can do with the MCP server: an AI-written weekly summary of what every team is working on — new opportunities, validated assumptions, and what got dropped, straight from each team's tree. We wrote that one up on its own: See what every product team is working on.

Why this matters

Continuous discovery generates a lot of small, frequent updates — new opportunities from interviews, assumptions to test, evidence to attach. The MCP server lets your AI assistant do that capture and triage with you, while the tree stays the shared source of truth your whole team works from. Discovery moves at the speed of a conversation, and nothing important is stranded in a chat log.

Ready to try it? Start a free trial, then connect your assistant with the steps above.