What is Cofound Agent?
Cofound Agent is a directory of people looking for a co-founder, built to be used by AI agents rather than people clicking through a website. You sign up, paste your LinkedIn profile or CV or upload a PDF, and check the structured profile it builds. After that there is no Cofound chat screen: you connect its MCP server to Claude, Claude Code, Cursor, ChatGPT, Codex or another MCP client and ask in plain words, such as "find me a technical co-founder in Europe open to fintech".
The server exposes 13 tools for your profile, search, messages and blocking. Your agent searches founders and people willing to join an idea, ranks matches by fit, writes first messages and drafts replies to pitches you receive, and asks you before anything is sent. Other users' agents see only broad categories: a role at Stripe becomes a top-tier fintech role, while names, exact companies, schools and dates stay on Cofound's server.
It is free with no card. The site does not name the company or people behind it.
Is Cofound Agent worth it?
Looking for a co-founder means hours of reading profiles on matching sites, most of them wrong for you, and answering pitches you should have ignored.
Cofound Agent hands that reading to the AI agent you already use. Once your profile is up, you ask Claude or Cursor to find, say, a technical partner open to fintech who can commit full time. Your agent searches, ranks people by how well your skills fit together, and drafts a first message. It also screens the pitches sent to you and asks "draft reply, or pass?" before anything goes out. Nothing sends without your confirmation.
The privacy idea is good. Other agents see your history only in broad buckets, so a stranger's agent learns that you held a senior role at a fintech firm, not which one or when. You decide when to reveal your name.
The catch: you are pasting your CV into a service that does not say who operates it. The privacy policy, effective 11 October 2026, names no company, address or AI provider, gives no retention periods and says nothing about training. A directory is only as useful as the people in it, and Cofound publishes no member count. Its GitHub plugin repository has no stars and no changes since 14 June 2026, and its tool reference caps first messages at 10 a day where the website says 20. The free-text parts of your profile are stored as you write them, so anything identifying you put there is visible.
Our take: cost is no reason to hold back, since it is free, but treat it like posting on a noticeboard owned by a stranger. Keep the profile to what you would put on a public page, and use it alongside, not instead of, the places where you already know who is on the other side.
Where it earns its place
A solo technical founder who works in Claude Code all day
Find a commercial co-founder without another website to check
Install the plugin with /plugin marketplace add cofound-agent/plugin, sign in once, and ask the agent each morning to search for business-side partners in your time zone and draft a first note; up to the daily new-contact cap, free.
A product manager open to joining a start-up but without an idea
Let founders come to me with ideas worth considering
Publish a willing-to-join profile and let other founders' agents pitch you; your agent filters by the criteria you set and offers a draft reply or a pass for each, so you read only the shortlist.
Before you commit
- Setup
- Sign up on the site with Google, GitHub or email, paste a LinkedIn profile or CV, check the profile, then add the MCP server at https://mcp.cofoundagent.ai/mcp (OAuth or a Bearer token) or the Claude Code plugin.
- Integrations
- 13 MCP tools covering profile, search, messaging and blocks; works with Claude, Claude Desktop, Claude Code, Cursor, ChatGPT, Codex, OpenClaw, Hermes and other MCP clients.
- Data and privacy
- Names, exact companies, schools and dates stay on Cofound's server and other agents see bucketed signals; free-text fields are stored as written. The privacy policy names no operator or AI provider, gives no retention periods and does not mention training; Google Analytics is used on the site. Which AI tools train on your data?
- Limits
- First messages to new contacts are capped per rolling 24 hours (20 on the site, 10 in the tool reference); search filters cannot use free text; no member count is published.
What’s good
- Free, with unlimited search and replies
- Works from Claude, Claude Code, Cursor, ChatGPT, Codex and any MCP client
- Other agents see only bucketed history, not names or employers
- Your agent asks before sending any message
What’s not
- Operator, location and AI providers not named
- No published member count, so match quality is unknown
- No data retention periods or training statement
- Message cap differs between site (20) and docs (10)
Best for
Founders who already work inside Claude, Cursor or another MCP client and want a low-effort, free extra channel for co-founder search with a deliberately sparse profile.
Cofound Agent: common questions
- Is Cofound Agent worth it?
- Cofound Agent is a clever, privacy-minded way to let your AI agent hunt for a co-founder, but nobody says who runs it or how many people are in it. We rate it 2.9 out of 5.
- How much does Cofound Agent cost?
- Free, no card. Unlimited search and replies in existing threads; the site caps first messages to new contacts at 20 per rolling 24 hours (the GitHub tool reference says 10) and full profile resubmissions at 20 a day.
- What are the downsides of Cofound Agent?
- Operator, location and AI providers not named; no published member count, so match quality is unknown; no data retention periods or training statement; message cap differs between site (20) and docs (10).
- Who is Cofound Agent best for?
- Founders who already work inside Claude, Cursor or another MCP client and want a low-effort, free extra channel for co-founder search with a deliberately sparse profile.
- Is Cofound Agent safe for company data?
- Names, exact companies, schools and dates stay on Cofound's server and other agents see bucketed signals; free-text fields are stored as written. The privacy policy names no operator or AI provider, gives no retention periods and does not mention training; Google Analytics is used on the site.
Answers checked Nov 11, 2026.