Pancake

• Tracking a competitor's LinkedIn posts and approving the people who engage with them for a short, signal-led outreach sequence
• Finding founders of recently funded startups that are hiring for a role your service supports, and getting new matches every morning
• Asking Claude or Codex to build a Play, show its credit cost, run it and summarize yesterday's leads best-fit first, without opening the Pancake app
• Running a separate Play for each audience an agency or consultancy sells to, all drawing on the same Brain
• Reviewing and approving new leads from Slack, and rejecting poor fits with a reason so the Brain's ideal customer profile improves

Pancake's strongest idea is the Play as a standing search with a reason attached to every lead. Instead of buying a database and building sequences, you describe an audience by what it is doing right now, and Pancake keeps surfacing people who match, shows its work, and waits for your yes before anyone is contacted. That makes it a good fit for founder-led sales on LinkedIn, and the MCP server is unusually well documented for driving it from Claude or Codex. The narrowness is the tradeoff: no email, no calls, no CRM sync, so it covers the top of the funnel and the first touch, not a full sales-engagement stack. Credits make the cost per lead variable, so start with the trial and one Play to see what a lead costs for your audience before scaling up, and remember that every message goes out from your own LinkedIn account.

About Pancake

Pancake starts by reading your website and building a GTM Brain: what you sell, your ideal customers, personas, competitors, messaging and voice. You then describe who you want to reach, in the Pancake web app or from an AI agent connected over MCP, and Pancake asks a clarifying question or two before saving a Play. A Play finds people one of eight ways: people engaging with posts by a competitor, an expert or your own brand, everyone who reacted to one post, people posting about chosen topics, companies hiring for a role, companies using a technology, recently funded companies, lookalikes of your best customers, or a straight role, industry, size and location match. Each search returns up to 50 leads, checks them against the Brain, and gives every lead one line on why it fits, drawing on 30+ data providers such as Dealroom, TheirStack, Coresignal and Wappalyzer that are included in the plan. The Play then runs on the schedule you pick, so new leads arrive in the app, in Slack or in a morning email digest. The agentic part has a clear boundary: searching is automatic, contact is not. You approve every lead, Pancake drafts a personal message from the lead's signal in your voice, and you can edit it before it goes out. Approved leads get a light warm-up and up to three LinkedIn messages, sent on weekdays during working hours from your own account (or rotated across up to ten), with a ten-minute undo window after enrolment. A reply stops the sequence and is labelled for you. Pancake is deliberately narrow: by its own statement of product boundaries, it does not send outreach by email, sync to a CRM, identify website visitors, look up phone numbers or make calls. Pricing is one self-serve plan from $99 a month per workspace with 60,000 credits, where credits pay for lead research and work-email lookups, plus a $3,000-a-month White Glove service and custom Enterprise pricing. The company was previously Basalt, an AI observability startup, announced the pivot to Pancake in July 2026 and relaunched it as "the Vercel for GTM" in October 2026, so expect the product to keep changing.