How to Pitch VCs from Claude
Paste one line into Claude, claude.ai, Claude Desktop, or Claude Code, and your agent can submit a venture pitch directly to Pitch Protocol's partner network. The line is: Read https://api.pitchprotocol.vc/skill.md to learn what Pitch Protocol is and how to use it. No separate web form, no account to create by hand first, no portal to dig through. The agent reads the skill file, learns the submission protocol, and handles the rest of the conversation with you.
This post is the step-by-step for doing that from inside Claude specifically: what to have ready before you start, how the connection actually works, what your agent submits, and what happens in the minutes and days after.
What to have ready before you start
You do not need a finished deck. Pitch Protocol's research pipeline runs fine without one, so if you have a working product, some early signal, and a clear sense of what the business does, that is enough to begin. What helps the session move faster is having the basics loaded into context your agent can draw from: a short description of the problem and the product, whatever traction you can point to, your cap table status, and the round size and terms you're thinking about. If you have a data room, a pitch deck, or even just a Notion doc with your numbers, point your agent at it before you start the submission so it has real material to package instead of guessing.
Connecting Claude to Pitch Protocol
Pasting the skill.md line is the primary way to connect, and it works the same way on every Claude surface, including Claude Code, which can fetch the URL directly inside a session. You do not need to register anything first. The skill file tells the agent which tools to call: it lists the relevant MCP tool names alongside the matching REST endpoints, so the agent knows exactly how to create an application, verify your email, and submit answers once the conversation gets there. Wiring Pitch Protocol in as a persistent connector is optional. If you want your agent to keep these tools available across sessions without re-reading the skill file each time, Claude's connector settings support adding remote MCP servers generally, the same mechanism used for any external tool. For Pitch Protocol specifically, the skill file is the path that works today, and it works on every surface, chat, desktop, or terminal, without extra setup.
What your agent actually submits
Once your agent reads the skill file, the sequence it follows is fixed and in order. First, it fetches the live application schema, so it knows exactly which fields Pitch Protocol expects right now rather than guessing from a cached format. Second, it creates the application, packaging your product description, traction, round details, and any documents you pointed it to. If you have a deck, it uploads as a PDF over plain HTTP at this stage, before email verification, so the file is attached to the application from the start. Third, it verifies your email: you receive a code, and the agent (or you, reading it off) confirms the submission is really yours. Fourth, a research pipeline runs, roughly five to ten minutes, checking what's publicly verifiable about the claims in your application. Fifth, the system returns tailored follow-up questions specific to your business, not a generic form, and your agent, with you in the loop, submits the answers. From there, partners across the network review the application, and the stated decision window is about 48 hours.
This is different from the agent "applying for you" in some opaque sense. The agent is doing exactly what you'd do by hand: fetching the form, filling in fields, verifying identity, answering questions, attaching context, except it's doing it in the same window where you've already been building, instead of making you context-switch to a browser tab and a fundraising CRM you've never used.
If you're building something agent-native, or you just do your best thinking inside Claude already, this is a reasonable way to find out if there's a thesis-aligned fund on the other end. Pitch Protocol's partner network spans more than 15 funds across pre-seed through Series B, and the application starts the same way every time: from your agent, with the skill file, in the tool you were already using.
The founder applies from their own agent. It's free. There is no web form.
Sources
- Connect to remote MCP Servers
- Connect to MCP servers, Claude Code Docs
- Pitch Protocol skill file
- Pitch Protocol
Skip the cold outreach. Apply once, free, from the AI agent you already use. It takes about ten minutes, and the funds that match reach out to you directly. Apply to Pitch Protocol →
Frequently Asked Questions
What exactly do I paste into Claude to start?
Read https://api.pitchprotocol.vc/skill.md to learn what Pitch Protocol is and how to use it. This works the same way on every Claude surface, including Claude Code.
Do I need a finished pitch deck before I start?
No. Pitch Protocol's research pipeline runs fine without one. A working product, some early signal, and a clear sense of what the business does is enough to begin, though having a data room, deck, or notes ready helps the session move faster.
What does the agent actually do after reading the skill file?
It fetches the live application schema, creates the application with your product description, traction, round details, and any documents, uploads a deck as a PDF if you have one, verifies your email with a code, waits roughly five to ten minutes for a research pipeline to check public claims, then submits answers to tailored follow-up questions. Partners then review, with a stated decision window of about 48 hours.