DiscoLike MCP Just Got the Feedback Loop

DiscoLike’s MCP just got the feedback loop. We turn an ICP prompt into target accounts and a clear query plan. Search terms are explicitly spelled out, no black box results.

Over half of our users are using Claude Code now. Another third are heavy API users. We are quietly powering many successful GTM automation companies.

A One-Way Street, Until Now

Until now, this was a one way street. Our clients ask and we deliver. In the background, our Outcome Learning Model converts an ICP prompt into a vector and keyword query, layers on tech stack, geo and other filters, and sends back target accounts.

MCP and API users can now insert themselves into the Outcome Learning Model loop. Use Claude Code with full client context. Adjust lookalike text, phrase matches and industry groups. Resubmit an updated query for better accuracy.

How the Loop Closes

Claude Code sends an ICP request to the DiscoLike search model. We send the query plan back to Claude for confirmation.

This closes the loop.

Both models iterate until balance is reached, then run full TAM queries and validation requests.

The Power of Feedback, From Humans to Models

Some of the best ideas come from your clients, all you need to do is listen carefully and spot the potential. The power of feedback, from humans to models.

Frequently Asked Questions

What is the DiscoLike MCP feedback loop?

The DiscoLike MCP feedback loop returns the explicit query plan, including vector text, phrase matches, and filters, back to Claude Code after every ICP request. Claude can adjust the plan with full client context and resubmit it. Both models iterate until the query is balanced before running the full TAM search.

Who can use the feedback loop?

Any DiscoLike user on the MCP or API. Over half of our users now run DiscoLike through Claude Code, and another third use the API directly. The feedback loop is available to both groups out of the box.

What is the Outcome Learning Model?

The Outcome Learning Model is the system behind DiscoLike that converts an ICP prompt into a vector and keyword query, layers on tech stack, geo, and other filters, and returns target accounts. It learns from thousands of successful TAM searches across hundreds of GTM teams.

Why expose the query plan instead of just returning results?

Black box results are hard to trust and hard to improve. Returning the explicit query plan lets Claude Code, with full knowledge of the client and the campaign, fine tune lookalike text, phrase matches, and industry groups before any large search runs. Better inputs produce better target accounts.

Related posts: