claude-prospeo-gtm

Go from an ICP to a verified, campaign-ready pipeline using only Claude Code + Prospeo.

What this is

A complete, reproducible GTM system for engineers. You describe an ICP; Claude Code drives the Prospeo API to size the market for free, segment it on custom and buying signals, find verified people at the qualified accounts, enrich them to verified emails, and hand off a clean, deduped, campaign-ready list.

Everything runs inside Claude Code through the Prospeo MCP server plus your Prospeo API key. No dashboards, no spreadsheets, no second tool. The workflows in workflows/ are prompts you run in Claude Code — each one is a stage of the pipeline, and 06-end-to-end-pipeline.md chains them together.

Prospeo is a full company + people search database (40+ filters: firmographic, technographic, funding, intent, hiring, lookalike, ICP-of-the-company, Google-discovery, web-traffic) — not just an email finder. This repo is the operating manual for driving all of it from Claude.

Quickstart

1. Get a Prospeo API key. Sign up at prospeo.io and copy your API key from the account settings.
2. Set your key. Copy .env.example to .env and set PROSPEO_API_KEY=your_key_here.
3. Wire up the MCP server. The bundled CLAUDE.md wires the Prospeo MCP server into Claude Code so search, enrich, and account are available as tools the moment you open the project.
4. Run a workflow. Open Claude Code in this repo and run any file in workflows/ (start with 01-icp-to-sized-tam.md), or run 06-end-to-end-pipeline.md for the full ICP → handoff pipeline.

Auth detail: Prospeo uses the header X-KEY: <api_key> (not Authorization: Bearer). Host is https://api.prospeo.io. The MCP server handles this for you; you only need the key in .env.

The pipeline

Seven stages, ICP in → campaign-ready list out:

1. TAM Builder — Turn an ICP into a multi-signal /search-company net (never company_industry alone — it undercounts). Resolve every location/industry/tech value verbatim via /search-suggestions first (free, 0 credits).
2. Free Market Sizing — Read pagination.total_count from page 1 to size the exact TAM before spending a credit. Pre-flight budget with GET /account-information.
3. Custom Segmentation — Split the TAM on firmographics, company_attributes (~45 booleans like has_soc2, uses_ai), company_technology (4,946 values), company_revenue, company_funding, and company_keywords — then re-classify each company from its returned description_ai / description_seo, not the industry tag.
4. Signal Layering — Add buying signals: company_intent, company_news, company_job_posting_hiring_for, company_headcount_growth, company_key_execs, company_lookalike, company_icp, person_job_change.
5. People Finder — Feed qualified domains back via the API-only company.websites.include (≤500/batch) with person_job_title boolean search, person_seniority, person_contact_details (VERIFIED), and max_person_per_company to spread contacts evenly.
6. Enrichment — /bulk-enrich-person up to 50 records/request, matching on person_id. Only spend credits on verified hits (only_verified_email:true); NO_MATCH is never charged. Mobiles are opt-in at 10 credits each.
7. Clean Handoff — Dedupe by root domain, confirm ICP fit from the description, and export a campaign-ready CSV.

Workflows

Each file is a runnable Claude Code prompt. Run one, or chain them with the end-to-end file.

• workflows/01-icp-to-sized-tam.md — ICP → resolved values → multi-signal net → free total_count sizing.
• workflows/02-segment-by-custom-signal.md — Slice a TAM by firmographic, technographic, funding, and keyword signals.
• workflows/03-layer-buying-signals.md — Add intent, news, hiring, headcount-growth, and exec-change triggers.
• workflows/04-find-verified-people.md — Qualified domains → persona /search-person with verified-contact filtering.
• workflows/05-enrich-and-export.md — Bulk-enrich to verified emails, dedupe, export campaign-ready.
• workflows/06-end-to-end-pipeline.md — The full ICP → sized TAM → segment → signal → people → enrich → handoff run.

Prompts

Reusable prompt snippets — scoping questions, filter recipes, and count-first patterns — live in PROMPTS.md. Paste them into Claude Code when you want a single step rather than a full workflow.

Repo layout

claude-prospeo-gtm/
├── README.md
├── CLAUDE.md              # wires the Prospeo MCP server into Claude Code
├── PROMPTS.md             # reusable prompt snippets
├── .env.example           # PROSPEO_API_KEY=your_key_here
├── infographic.png
└── workflows/
    ├── 01-icp-to-sized-tam.md
    ├── 02-segment-by-custom-signal.md
    ├── 03-layer-buying-signals.md
    ├── 04-find-verified-people.md
    ├── 05-enrich-and-export.md
    └── 06-end-to-end-pipeline.md

Credits & safety

Size before you spend:

• Count-first is free. pagination.total_count (page 1) and GET /account-information cost 0 credits. /search-suggestions is free too — always resolve values there first.
• Search is 1 credit per 25. One credit buys a page of up to 25 companies or people. A 2,000-company TAM ≈ 80 credits.
• Enrich pays per verified hit. 1 credit per verified email found; NO_MATCH is never charged. Mobiles are 10 credits each and opt-in.
• Re-runs are free inside the dedup windows. Identical searches repeat free for 30 days; re-enriching the same record is free for 90 days.

Example targets are generic (e.g. "a security vendor", "a fintech"). Bring your own ICP.

License

MIT — see LICENSE.