Setup
Upload your LinkedIn export once — the same "Request my data" zip LinkedIn already gives you. PersonalKnowHow builds your knowledge graph from it and hands you back a link. Add that link as a connector in Claude, or any MCP-compatible assistant, whenever you want to use it — that part's on you. (GitHub, course platforms, and more sources are on the roadmap — not part of the hosted upload yet.)
Usage — match against a job posting
"Would I be a good candidate for this job? [link to a real posting]"
It reads the posting itself and checks it against your real evidence:
✓ Snowflake pipelines — real production project
✓ dbt — course completion + hands-on project
✓ GDPR-compliant data handling — real project, not a course
⚠ Airflow — course completion only, no production use yet
Just the link. No summarizing the posting yourself first.Usage — tailor your CV
"Tailor my CV for this job."
Every bullet traces back to something real — a project you shipped,
a course you finished, work you actually did. Not invented phrasing.How it works
-
1
Collect
Upload your LinkedIn export — the "Request my data" zip from LinkedIn's own Settings & Privacy. That's what the hosted product ingests today; more sources are on the roadmap.
-
2
Understand
Recognizes the skills and tools behind what you've actually done, not just the exact words you used to describe them. Real AI embeddings, not keyword tags — works the same across any field.
-
3
Connect
Links everything into one graph, with duplicates removed — safe to rebuild any time you add more.
-
4
Answer
Your AI can ask it anything, anytime, and get back real answers — ranked by what's actually relevant, not just matching words.
Try the live demo
https://personalknowhow-demo.kxtwrdzt6g.workers.dev/mcp is a real, deployed
MCP server — but it's not a webpage. Opening that URL in a browser sends a plain
GET, and MCP only speaks POST with JSON-RPC framing, so you'll
just see a bare error. That's expected — it means you're looking at it the wrong way.
The real way to use it: add it as an MCP connector in Claude Desktop.
{
"mcpServers": {
"personalknowhow-demo": {
"command": "npx",
"args": ["mcp-remote", "https://personalknowhow-demo.kxtwrdzt6g.workers.dev/mcp"]
}
}
}Restart Claude Desktop, then ask it to check whether "you" have experience with any
skill — Claude calls the query_knowhow tool over MCP and returns
semantically-matched evidence with similarity scores, no auth required.
Prefer a quick terminal check instead of a full MCP client?
curl -s https://personalknowhow-demo.kxtwrdzt6g.workers.dev/mcp \
-X POST -H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}}}'A 200 with a JSON-RPC response back confirms it's live.
Privacy by construction, not by promise
The public demo above is backed by a fail-closed allowlist — only explicitly listed categories (courses, projects, certifications, education, endorsements, positions, recommendations, articles) are ever exported to it. A new data category is excluded by default until someone deliberately adds it. Anything more sensitive — job applications, career interests — lives only in a separate, bearer-token-gated server with no shared code path to the public one, and is never committed to the public repo at all.
Want your own?
Right now this is a personal project you can fork and self-host. We're gauging interest in a hosted version — bring your own LinkedIn/GitHub, get your own graph and MCP server, no setup required. No promises on timeline or feasibility yet — if that's something you'd actually use, tell us.