We turned Claude into a VC outreach machine
The exact setup behind 400+ VCs in our pipeline on day one, from the context Claude reads first to the approval step that keeps a person on every send.
Free guideSetup in about 45 minContents
What this is
Fundraising research is the slowest part of a raise. Weeks of spreadsheets, half-read portfolio pages, and cold notes that read like they came off the same template.
We handed that work to Claude. On day one of pressing go, 400+ VCs were in our pipeline: researched, ranked against our thesis, conflicts removed, each with a first note drafted in our voice and waiting for approval.
This guide is the whole setup, step by step. You can rebuild it this afternoon.
What you’ll have at the end
- A ranked database of funds that match your stage, check size and thesis
- Funds that backed a direct competitor, removed before you write to them
- A “why them” line for every fund, linked to the deal or post it came from
- First notes drafted in your voice, queued for your approval
- A follow-up cadence, and one place to log every reply
Who it’s for
Founders raising pre-seed to Series A who run their own investor outreach and already use Claude. The same steps work for angels and accelerators.
Why Claude alone writes the same email 400 times
Ask Claude for a VC list and a first note and you get a reasonable list and a note any founder could have sent. Every partner has read that note. Probably this morning.
What’s missing is context. Claude starts every chat knowing nothing about your company: no deck, no metrics, no sense of which funds are off limits, no idea how you write. You can paste it in, but the paste is partial, it’s gone by the next chat, and 400 notes later the details have drifted.
So we gave Claude a brain to read before every prompt. One place that holds the deck, the traction, our investor criteria, our competitor map and our voice, connected to Claude over MCP. Every prompt starts from what’s true about the raise, and every claim in a draft points back to a document we wrote.
Your documents
- Company brief
- Traction
- Investor profile
- Competitor map
- Voice guide
Shared context
HeyBrainPermissions, citations, access recordResearch and writing
ClaudeFinds, ranks and draftsEvery send
Four mistakes this setup removes
- Writing to a fund that backed your competitor. It’s easy to miss on a portfolio page, and it ends the conversation. Every fund is checked against your competitor map first.
- Personalisation that stops at a first name. Each note opens with a reason tied to a real deal, thesis post or partner focus, and cites it.
- Stage and check-size mismatch. A $1.5M seed round doesn’t belong in a growth fund’s inbox. Your investor profile filters these out before anything is ranked.
- Sending everything at once. Notes queue for your review and go out in daily batches from your own inbox.
What you need
The setup unlocks with your email
Steps 1 to 6 and every prompt are below. Add your email once and the whole guide stays open.
Cursor
CodexWhy this runs on HeyBrain
Every step above depends on one thing: Claude reading the same, current context before it writes. HeyBrain is built to be that context.
- One brain for every AI tool. The same context works in Claude, Cursor and Codex over MCP, so a cofounder or an agent reads exactly what you read.
- Scoped access. Review the access you grant each connected AI tool. Manual uploads have their own Brain permissions; they do not inherit source-folder permissions.
- Review the sources. Ask Claude to cite its source documents and check those citations before using a number or sending a draft.
- Your data stays yours. Nothing you connect is used to train any model of ours.
The free plan covers this whole guide. After the raise, the same brain keeps working for investor updates, hiring and sales.
Questions founders ask
No. Free Claude plans can add one custom connector, which is all this guide needs. Paid plans allow more connectors, and longer research runs go faster with higher usage limits.
There’s a free plan with a monthly allowance of credits, spent on queries and on indexing new content. It covers this guide. Paid plans raise the allowance and add more sources, seats and agents.
No. HeyBrain uses governed retrieval only, and nothing you connect is used to train any model of ours.
No. HeyBrain holds your context and Claude drafts. You approve every note and send it from your own inbox or sequencing tool.
HeyBrain connects to Claude, Cursor and Codex over MCP today. The prompts here are written for Claude.
It’s a strong first pass with a source on every line, which is what makes it checkable. Funds change partners and pause investing, so spot-check the top of your list before you send.
Not when each one is written to one partner, gives a real reason, is approved by you, and goes out in small daily batches from your own inbox. Keep it that way.
