---
title: "revops-meddicc-agent: GitHub Repository"
url: https://stacklist.com/card/6eb6ec23-0175-499e-9131-02154a3ebf2b
source_url: "https://github.com/jeff266/revops-meddicc-agent"
stack: https://stacklist.com/c/technology/stack/cc483fc9-ce97-46ea-bc49-7a300af7cd39
summary: "RevOps MEDDICC Agent is an AI-powered tool that automatically analyzes sales call transcripts against qualification methodologies and maintains accurate deal scoring in HubSpot every night. It provides pipeline analytics including waterfall tracking, win/loss narratives, objection logging, and feature-gap analysis to improve sales forecasting and team alignment."
tags: "revops, sales-automation, meddicc, hubspot, call-intelligence, pipeline-management, ai-agent"
key_entities: "RevOps MEDDICC Agent (technology), HubSpot (technology), Claude Code (technology), GitHub Actions (technology), Supabase (technology), Fireflies (technology), Gong (technology), MEDDICC (concept), MEDDPICC (concept), SPICED (concept), BANT (concept), Jeff (person), RevOps Impact (organization)"
classification: "reference"
content_hash: "sha256:f524d170229ac17fe99fcafc1b119a48addd13c2ece8a01d25f573d4aeb61d9e"
acp_version: "0.2"
token_counts_approximate: 1395
visibility: public
agent_accessible: true
status: "final"
---

# revops-meddicc-agent: GitHub Repository

RevOps MEDDICC Agent Your sales calls contain the truth about every deal's health, but that truth usually dies in a rep's head or a forgotten Gong recording. This agent listens to every call, reads it against your qualification methodology, and keeps every deal in HubSpot honestly scored, every night, without anyone doing manual review. Each night it re-evaluates your full active pipeline. It pulls the latest call transcripts, scores every deal on MEDDICC (or MEDDPICC, SPICED, or BANT, configurable to how your team actually sells), and writes the scores straight back to HubSpot so reps and managers see the same picture. It flags risk before it becomes a surprise in the forecast call, and it gets sharper over time as it learns your team's patterns. Beyond nightly scoring, it runs a weekly analytics pass across the whole pipeline: Waterfall tracking : see exactly where deals are moving forward, sliding back, or dying, stage by stage Win/loss narratives : automatically extracted from call evidence, not just the close reason field Objection log : a searchable record of what prospects actually pushed back on Feature-gap backlog : requested features pulled straight from calls, ranked by how often and how severely they come up Interested in forward deployment for your team? Reach out: Email: jeff@revopsimpact.com LinkedIn: linkedin.com/in/jeffbethechange Setup: three steps Step 1 · Credentials Open this repo in Claude Code. If this is a fresh fork, Claude Code will detect the missing config files and guide you through setup automatically. Just open the project and say: "set up this repo" Claude Code will: Walk through every credential and API key Write config/client.yaml and config/context.yaml Discover your HubSpot stage IDs Set up Supabase Verify everything before the first run Step 2 · Add GitHub Secrets After Claude Code generates your .env file, add the values as GitHub Secrets: repo → Settings → Environments → Agent → Add Secret Or with the GitHub CLI: gh secret set --env Agent --env-file .env Step 3 · First run Go to: Actions → MEDDICC Agent Nightly Run → Run workflow Watch the logs. First run analyzes your full active pipeline. After that the agent runs every night at 2am UTC automatically. What runs automatically Time (UTC) Job What it does 1:00 AM Daily Deal ETL Updates active deal index from HubSpot 1:30 AM Daily Calls ETL Fetches new calls for active deals 2:00 AM MEDDICC Agent Analyzes deals, writes to HubSpot + Supabase Call intelligence platforms The agent supports two call recording platforms: Fireflies (default): Fireflies.ai call transcripts Most common for SMB/mid-market Simple API key authentication Set call_tools.primary: "fireflies" in config/client.yaml Gong : Gong.io enterprise call intelligence Enterprise standard for larger sales teams Provides richer structured data (topics, action items, talk time) Requires Access Key + Access Key Secret Set call_tools.primary: "gong" in config/client.yaml Claude Code will automatically detect your choice during setup and collect the right credentials. What runs nightly 2am UTC: GitHub Actions fires → Load active deals from deal index → For each deal: load call cache → context builder (Haiku) → Generator (Sonnet) → Evaluator (Haiku) → Reflection gate → Write analysis to GitHub output/ → Write 6 MEDDICC scores to HubSpot deal properties → Write analysis to Supabase for query layer → Update CLAUDE.md via PR if new patterns emerge What runs weekly (analytics workflow) Sundays 3am UTC: GitHub Actions fires 1. Analytics deal ETL: fetch all deals (all stages, both analyzed and analyze:false pipelines) 2. Snapshot deals: capture point-in-time pipeline state to deals_snapshot table 3. Compute waterfall: track qualified pipeline movement across 5 categories (new, newly qualified, forward, backward, won, lost) with reconciliation check 4. Generate win/loss narratives: extract call evidence and compare to stated close reasons 5. Extract objections: categorize and store per-company objections from call transcripts 6. Extract feature gaps: identify and severity-score requested features from calls Output: Waterfall movements in waterfall_weekly table, win/loss patterns, objection vault, and feature request backlog. All queryable via Supabase or the CRO agent. Files to know File What it does scripts/run_nightly.py Main orchestration, runs every night scripts/meddicc_agent.py Generator + evaluator + reflection loop scripts/etl_calls.py Builds call cache from CSV exports scripts/etl_deals.py Builds deal index from HubSpot prompts/CLAUDE.md Generator instructions, edit to calibrate prompts/evaluator_rubric.md Evaluation criteria, auto-improves config/client.yaml Your HubSpot stage IDs and thresholds config/context.yaml Your competitors, objections, feature gaps memory/calls/ Call cache, 1 JSON per company memory/learnings/ What the agent is learning output/ MEDDICC analysis files Costs Scenario Cost First full pipeline run ~$3-5 Nightly steady state ~$0.10-0.30 Monthly total ~$10-15 Skills for Claude.ai and Claude Code This repo includes two onboarding skills: skills/revops-agent-setup/SKILL.md : credential setup wizard skills/revops-client-context/SKILL.md : client context onboarding In Claude Code (desktop app): The skills run automatically when you open a fresh fork (missing config files). Claude Code will proactively offer to guide you through setup. In Claude.ai (web/mobile): Copy the contents of each SKILL.md file into Claude.ai's custom skill creator. Once saved, trigger by saying "start client onboarding" or "set up credentials".