Find
Headhunt, prospect, or map a market. Ask in plain English and get back a real shortlist - people, companies, repos - read live from the open web.
A sentence in. The agent reads the open web, live.
- Results will appear here as the agent finds them.
How it works
One sentence in. A shortlist out.
- 01
Describe who you're after
Plain English, no filters or booleans. "Staff backend engineers in NYC who shipped Rust in production" is a complete query.
Try that exact search - 02
I read the open web, live
I plan the search, then read public LinkedIn and Crunchbase records, GitHub, and the sources they link to - at search time, not from a stale export. You watch every step as it runs.
- 03
Act on a ranked shortlist
Every result comes with the reasoning behind it. From there I draft outreach in each person's language, or you export the list to wherever you work.
See real shared results
Trusted by
Individual recruiters, investors, and founders.
The ones who source for themselves - from teams like these.
Ramp
Mercury
Linear
Retool
Vanta
Deel
Hunt Club
Riviera Partners
Uncork Capital
Craft Ventures
Boldstart
Amplify Partners
Base10
Initialized
Reach each one, in their language, with the line that earned them in.
Sarah K.·San Franciscomaintains tikv/raft-rs · 314 commits in last year
drafting
Mike T.·San Franciscoex-Stripe · founding eng at Quantum Forge
queued
Emily H.·New YorkRustConf ’24 · "async runtimes that don’t lie"
queued
Chris P.·San Franciscoex-Apple · founding eng at Foundry (18 ppl)
queued
You stay in the loop on every send.
API and MCP server
Put people search in your own stack.
Everything the interface does is one HTTP call away, and the same searches drop into Claude, Cursor, or VS Code as tools your assistant can call mid-conversation. One key, the same credits, no separate plan.
Call it from your code
A sentence in for a ranked shortlist with the reasoning attached, or an exact filter set for the same page every time.
curl https://www.refolk.ai/api/v1/search \
-H "Authorization: Bearer $REFOLK_API_KEY" \
-H "Content-Type: application/json" \
-d '{"query": "Staff backend engineers in NYC who shipped Rust in production"}'Or add it to your assistant
One URL, one header, 4 tools. Then ask for a shortlist in the chat you are already in.
claude mcp add --transport http refolk https://www.refolk.ai/api/mcp \
--header "Authorization: Bearer rfk_live_your_key_here"A search in plain English costs 100 credits, and nothing at all when it finds nobody; a filtered search costs 10. A new account starts with 500 free credits, so the first calls are on me.
Shared searches
See what others are looking for.
Real prompts, real results. Each page is a snapshot of a search someone ran and decided to share.
- Shared
Senior Rust engineers in the US with public GitHub activity in the last 12 mont…
- Shared
Show me engineers in Berlin or Munich with Rust experience who have public comm…
- Shared
Engineers in the US who list Apache Airflow as a skill and have committed to it…
- Shared
US senior software engineers with production Ruby on Rails experience and eithe…
- Shared
Open-source developers who have contributed to more than one competing agent fr…
- Shared
Find US-based staff or principal backend engineers with 9+ years of experience …
For developers
The same search, from your own code.
Everything the interface does is an endpoint: a sentence in, a ranked shortlist with the evidence out, or an exact filter set in and a deterministic page out. Point an assistant at /api/mcp and it can do the same thing mid-conversation.
curl https://www.refolk.ai/api/v1/search \
-H "Authorization: Bearer $REFOLK_API_KEY" \
-d '{"query": "Staff Rust engineers in NYC"}'No SDK to install. Read the API documentation.
From the blog
Playbooks and field notes.
Headhunting, sourcing, and prospecting - written for people who actually have to do it.
- 10 min read
LinkedIn's 4.77% Engineer Reply Rate Is Below Its Own 13% Floor
LinkedIn's software InMail reply rate is 4.77%, a third of its own 13% enforcement floor, while overage credits hit $21. Here is the math and the exit.
Read post → - 9 min read
EU AI Act Hiring Split: 831 Recruiters Per Governance Hire
The Digital Omnibus deferred Annex III to December 2027. Why CV screeners face €15M fines while outbound sourcing can still fit through Article 6(3).
Read post → - 10 min read
216 Applications, 10 Offers: The Ceiling on HN's Best Inbound
One HN Who Is Hiring employer disclosed 216 applications and 10 offers. The 4.6% math explains why founders still need outbound sourcing.
Read post →
Questions
The short version.
What is Refolk?
Refolk is an AI agent for headhunting and prospecting. Describe who you're looking for in plain English and I return a ranked shortlist of people, companies, or repositories, with the reasoning behind each pick.
What can I search for?
People - engineers, founders, operators, investors. Companies - by industry, stage, geography, or investor. GitHub repositories and the maintainers behind them. If it's on the open web, I can probably find it.
Who is it for?
Anyone whose job starts with finding a specific person or company. Headhunting and sourcing, sales and partnership prospecting, investor deal flow, market and competitor research. It is the same search either way: describe who you want, read the shortlist, see why each name is on it.
Where does the data come from?
Public LinkedIn and Crunchbase records, GitHub, and the open web - read live at search time, not from a stale database export.
What does it cost?
You can start for nothing: 500 free credits, no card required, and a balance that refills every month for as long as your account is open. Paid plans raise that monthly number when the free one stops being enough.
Do I need a credit card?
No. Signing in with your email is enough. Credits are granted to your account directly.
Is there an API?
Yes, and an MCP server with it. The same two searches run from your own code over one HTTP endpoint and one key, or from inside Claude, Cursor, or VS Code as tools your assistant can call mid-conversation. Same account, same credits, no separate API plan.
How is this different from LinkedIn or GitHub search?
Those tools need exact filters and keywords. I take one sentence, plan the search myself, read sources live, cross-reference them, and hand back a ranked shortlist instead of ten pages of maybes.
Can I use it from my own code?
Yes. Both searches are a keyed HTTP API, and there is an MCP server so Claude, Cursor, VS Code, or anything else that speaks the protocol can run them mid-conversation. Same account, same credits, documented in full at /developers.
More detail on the pricing page, and in the API documentation.
Find anyone. Just ask.
Your first searches are on me - 500 free credits, no card required.