Skip to main content
View .mdOpen in ClaudeOpen in ChatGPT

For AI agents

These docs are designed to be consumed by AI coding agents — Claude Code, Cursor, Windsurf, ChatGPT, Copilot, Gemini, anything — as first-class readers alongside humans. If you're an agent reading this, you're in the right place. If you're a human thinking about how to use your agent with FunnelFizz, this is also for you.

The three agent-facing artifacts​

1. /skill.md — the single-file setup bundle​

URL: https://docs.funnelfizz.com/skill.md

Drop this URL into your agent and say "follow this to set up FunnelFizz." It's a 2,500-word self-contained runbook that teaches the agent to walk you through signup → tracking install → Stripe connect → provider connect → sending domain → identify + custom events → first automation.

The file is designed to be pasted whole into an agent's context. No cross-page dependencies, no navigation chrome, no boilerplate. Just instructions.

Inspired by Moltbook's skill.md pattern.

→ Full walkthrough of how to hand it to your agent

2. /llms.txt and /llms-full.txt — the docs corpus​

URLs:

Following the llms.txt convention proposed by Jeremy Howard / Answer.AI, we emit both a sparse index (llms.txt) and a dense dump (llms-full.txt).

Use llms.txt when your agent only needs a map — "which page should I read about splits?"

Use llms-full.txt when your agent needs the whole corpus in context — "answer general questions about FunnelFizz."

Both files are regenerated on every docs build so they're always current.

→ Details on the llms.txt convention and when to use which

3. Per-page .md routes — clean markdown for any page​

Pattern: any page URL + .md → the markdown source, minus navigation and footer chrome.

Examples:

  • Page: https://docs.funnelfizz.com/concepts/stages
  • Markdown: https://docs.funnelfizz.com/concepts/stages.md

These are auto-generated from the source MDX and stripped of all HTML wrapping. They're the cleanest thing to paste into an LLM when you want it to answer a question about a specific topic without burning tokens on site chrome.

The "contextual menu" on every page​

At the top of every docs page (after the title), you'll see a small row of buttons:

  • Copy as markdown — copies the clean .md version to your clipboard.
  • Open in Claude — opens claude.ai with the page preloaded as context.
  • Open in ChatGPT — same, for chat.openai.com.
  • Copy MCP URL — copies the FunnelFizz MCP server URL (see below).

(These are added via a swizzled DocItem footer component; if you don't see them yet, they may not be shipped in your build.)

MCP server​

We ship a Model Context Protocol server that exposes read-only FunnelFizz state to compatible agents (Claude Desktop, Cursor, Windsurf, etc.).

Available tools (read-only):

  • List your funnels
  • Get a funnel's current metrics (counts + rates per stage)
  • Get a specific profile's journey timeline
  • Check tracking site verification status
  • Get Stripe integration health

Install: see MCP & Context7 for the one-line install.

Context7​

FunnelFizz is indexed in Context7, so agents that support Context7 MCP can fetch our docs by saying "use context7 funnelfizz". Our context7.json lives at the root of our docs repo.

→ MCP & Context7 setup

Principles​

We try to make these docs AI-agent-friendly not just by publishing machine-readable formats but by writing for both audiences at once:

  • Use absolute URLs in examples, not relative links that break on copy-paste.
  • Quote exact enum values. If we say the split condition is traffic_source, that's the literal string. Agents can trust it.
  • Avoid ambiguous pronouns ("it", "this") without a clear antecedent.
  • Put the conclusion first. TL;DRs at the top, explanations after. Agents scan before they read.
  • Mark preview features explicitly. If something's not shipped ("SEO dashboard is demo-only today"), we say so. Agents that commit to non-existent features cause real user pain.

For humans thinking about this​

If you're a solo founder wondering "wait, do I really need to think about AI agents reading my funnel setup docs?" — probably not directly. The reason this matters is that your users will increasingly delegate tool-setup to their agents. Making FunnelFizz easy to set up by agent is the same thing as making FunnelFizz easy to set up for busy humans. Agent-friendly = human- friendly.