Playcall

Playcall is an MIT-licensed alternative to Gong for teams that want their call scoring to be inspectable and self-hosted. The distinction it draws is that most call intelligence grades the transcript in isolation; Playcall grades it against the account, the contact's role and the deal stage, so the same discovery question can score well on one call and badly on another.

You give it a methodology — MEDDPICC, BANT, SPIN or a custom playbook written in your own words — and it generates the scoring rubric, then returns evidence-backed scores per rep with the passage behind each judgement. Beyond per-call grades it surfaces recurring objection patterns across a set of calls, ties loss reasons to what was actually said rather than to a CRM dropdown, and pairs each gap with a coaching drill.

The stack is deliberately ordinary and cheap to run: Next.js 16 with React 19, Supabase for Postgres, auth, row-level security and edge functions, Vercel Blob for uploads, OpenAI Whisper for transcription and LlamaParse for documents. Model choice is bring-your-own-key per workspace through the Vercel AI SDK, covering OpenAI, Anthropic, Gemini, Mistral, Groq, Cohere, Perplexity and Together AI, with Exa or TheHog for buyer and company enrichment.

The software is free and the source is on GitHub; a self-hosted deployment on Vercel and Supabase runs at roughly $50/month of infrastructure plus whatever your model and enrichment calls cost. That makes it a realistic option for a small GTM team that would never clear the budget or the procurement cycle for an enterprise revenue-intelligence contract.