PMM Sherpa MCP

Senior product marketing judgment, in every AI you use.
One MCP server. 38,000+ curated chunks of PMM wisdom. Four tools. Works across Claude.ai, Claude Code, ChatGPT, Codex, Gemini CLI, and Antigravity.

🏔️ Try Sherpa on the web  |  📖 Full Docs  |  💬 Discussions

You're staring at a positioning doc, a launch plan, or a battlecard. You know the framework. You can't decide if you're applying it right.

PMM Sherpa is the senior PMM you wish you could DM at 11pm. It is a remote MCP server grounded in 38,000+ curated chunks from PMM books, AMAs, podcasts, and practitioner essays, with a calibrated advisory voice and four purpose-built tools. Connect it to whichever AI client you already use, and PMM judgment shows up in your normal flow, not in a new tab.

✨ The four tools

| Tool | When to use it | What it returns | |---|---|---| | ask_sherpa | Strategic framing, "is this right?", "how should I think about X?", open-ended PMM judgment | Advisory dialogue grounded in the corpus, in Layer 4 voice | | draft_artifact | Generating one of 39 named PMM deliverables (positioning, messaging framework, ICP, persona, battlecard, landing page copy, launch plan, demo script, ad variants, comparison matrix, case study, analyst briefing, board deck, and more) | Structured deliverable with framework baked in | | get_feedback | Pressure-testing a draft, landing page, deck, or copy you already have | Critique against senior PMM heuristics: gaps + recommendations | | scope_pmm_research | Planning phase of a Deep Research run on a PMM-adjacent question | Research brief: angle, sub-questions, sources to weight, anti-patterns |

All four are grounded in the same corpus and speak the same voice. You orchestrate. Sherpa supplies the judgment.

Quick Start

Pick your client. Each setup takes a minute.

 Claude.ai  ·  Pro / Max / Team / Enterprise

1. Claude.ai → Settings → Connectors → Add custom connector
2. Name: PMM Sherpa  ·  URL: https://pmmsherpa.com/api/mcp  ·  Auth: OAuth
3. Sign in with the same Google account you use on pmmsherpa.com
4. (Recommended) Upload the Claude.ai skill so Sherpa auto-triggers on PMM keywords with the right voice
5. Ask: "Help me sharpen positioning for [your product]."

 Claude Code  ·  CLI + IDE extensions

claude mcp add pmm-sherpa --transport http https://pmmsherpa.com/api/mcp

OAuth opens in your browser on first call. Then install the skill so Sherpa auto-fires on PMM keywords:

git clone https://github.com/boommark/pmmsherpa-mcp.git ~/.pmm-sherpa
mkdir -p ~/.claude/skills
ln -s ~/.pmm-sherpa/skills/claude-code ~/.claude/skills/pmm-sherpa

In your next Claude Code session, prompt:

> See pmmsherpa.com. Help me create a battlecard against its top competitor.

 ChatGPT  ·  Plus / Pro / Team / Business — Developer Mode

1. ChatGPT → Settings → Connectors → Advanced → toggle Developer Mode ON
2. Settings → Connectors → Create:
   • Name: PMM Sherpa
   • MCP Server URL: https://pmmsherpa.com/api/mcp
   • Authentication: OAuth, "I trust this application"
3. Sign in with the same Google account as pmmsherpa.com
4. Settings → Personalization → Custom Instructions → paste skills/chatgpt/custom-instructions.md (~1,200 chars). This is the skill-equivalent for ChatGPT, since it does not yet support skills natively.
5. In a new chat, toggle PMM Sherpa ON in the Tools panel. Ask: "Help me sharpen the positioning for [your product] against [competitor]."

Codex CLI, Gemini CLI, and Antigravity also work. See Multi-client install below.

🌐 Multi-client install

Sherpa works in every modern MCP-capable AI client. Pick yours.

Help me build a battlecard against [competitor] for [your product].

claude mcp add pmm-sherpa --transport http https://pmmsherpa.com/api/mcp

git clone https://github.com/boommark/pmmsherpa-mcp.git ~/.pmm-sherpa
mkdir -p ~/.claude/skills
ln -s ~/.pmm-sherpa/skills/claude-code ~/.claude/skills/pmm-sherpa

> /mcp
> See pmmsherpa.com. Help me create a battlecard against its top competitor.

Help me sharpen the positioning for [your product] against [competitor].

codex mcp add pmm-sherpa --transport http https://pmmsherpa.com/api/mcp

{
  "mcpServers": {
    "pmm-sherpa": {
      "transport": "http",
      "url": "https://pmmsherpa.com/api/mcp"
    }
  }
}

📦 Skills + custom instructions

Sherpa speaks a calibrated voice (Layer 4: discovery cadence, story-first, framework-named-mid, single-question close, never em-dashes). The skills enforce this voice and auto-route to the right tool.

| Client | Format | File | |---|---|---| | Claude.ai | Skill (zip) | skills/claude-ai/pmm-sherpa-skill.zip | | Claude Code | Skill folder | skills/claude-code/ | | ChatGPT | Custom Instructions text | skills/chatgpt/custom-instructions.md | | Codex / Gemini CLI / Antigravity | Use the ChatGPT-style custom instructions | Same file as ChatGPT |

Without the skill, Sherpa still works. You just have to remember to toggle the connector ON and pick the right tool. With the skill, it picks correctly on its own.

💡 Examples

Five sample workflows, end-to-end:

• Prep for a PMM interview at a target company. ask_sherpa scopes the company, draft_artifact builds a teardown, get_feedback pressure-tests your case answers.
• Positioning sharpening in Claude.ai. From rough draft to crisp positioning statement, with one Sherpa call.
• Launch deck via skill in Claude Code. Skill auto-fires, draft_artifact produces the deck outline.
• Battlecard in ChatGPT. Custom instructions route to draft_artifact, response stays in voice.
• Deep Research with Sherpa. scope_pmm_research plans, subagents call ask_sherpa, get_feedback synthesizes.

💰 Pricing

| Plan | Monthly | What you get | |---|---|---| | Free | $0 | 10 MCP credits / month. 2 credits per query. Try every tool. | | Starter | $9.99 | 200 MCP credits / month. Unlimited use of pmmsherpa.com chat. Topup packs unlocked. |

Topup packs (one-time, never expire, Starter only):

| Pack | Credits | $/credit | |---|---|---| | $5 | 50 | $0.10 | | $10 | 125 | $0.08 | | $15 | 200 | $0.075 (best value) |

See full pricing  ·  Purchase docs

📚 The corpus

38,000+ chunks, hand-curated across the highest-signal PMM material in the field:

• 34 authoritative books on positioning, messaging, pricing, GTM, and launch
• 1,100+ practitioner conversations: podcast episodes and senior-PMM AMAs
• 1,600+ practitioner essays: long-form blog posts and operator write-ups

The corpus is curated, not crawled. Every source is hand-picked for substance density. Want something added? Open a corpus request.

🎙️ Voice & cadence

Sherpa speaks in Layer 4 voice. Every response:

• Opens with your situation, not preamble. No "Great question."
• Tells the story before naming the framework.
• Varies paragraph length. Long build. Short punch. Question. Space.
• Closes with one question you take away. Never a menu of options.
• Never uses em-dashes. Period, comma, colon, or parentheses instead.
• References principles. Never names the authors, books, or podcasts in the corpus.
• Doesn't hedge ("it depends", "could be") when the corpus has a clear position.

Same voice senior PMMs use in private Slack DMs. It transfers if you install the skill.

❓ FAQ

🤝 Contributing

This repo is the public docs and skill mirror. The Sherpa server itself is closed-source. We welcome:

• Bug reports and feature requests for the MCP server, skills, or docs (open an issue)
• Corpus requests, books, AMAs, podcasts you want added (open a corpus request)
• Doc PRs for the skill files, examples, or README

See CONTRIBUTING.md for the full guide.

📜 License

MIT for the docs, skills, and examples in this repo. The Sherpa server, corpus, and proprietary prompts are not open-sourced.

🙏 Acknowledgements

• The PMM community whose books, AMAs, podcasts, and essays form the corpus
• The MCP spec authors at Anthropic for designing the protocol that makes this possible
• The Vercel AI SDK team for the streaming + tool-calling primitives
• The Supabase team for pgvector
• Every PMM who beta-tested Sherpa and pushed back on bad outputs

Built by DBAR LLC. Sherpa speaks because senior PMMs bothered to write things down.