Stop pasting job descriptions into your AI.
jd-intel gives your assistant a direct line to real, current job openings. You ask in plain language. It finds the roles that fit you and ranks them against your background. No copy and paste, no new tool to learn.
Everyone says start with the job boards. Both ways of doing it cost you.
However you chase an edge in the search today, you pay a tax on your own leverage.
You become the data clerk.
Copy a job description, paste it in, write a fresh prompt, repeat for the next role. None of the effort carries to the next search. The work is yours, every time.
A black box ranks you.
Their algorithm decides what surfaces, tuned for their goals. You never see why a role appeared, or which ones never did. The judgment is theirs, not yours.
jd-intel takes the tax off. It hands current, real postings to the assistant that already knows your background, so your context does the ranking. Not a board's algorithm. Not another tab to manage.
Set it up once. Then you just ask.
Set it up in your AI assistant
One file in Claude Desktop. No terminal, no Node.js, about two minutes. Other MCP clients take one small config.
Ask in plain language
"Find product roles at fintechs posted this month." No filters to configure, no syntax to memorize. Just say what you are looking for.
It answers with real roles
Current openings, pulled the moment you ask and ranked to you, with links to apply. Then keep going: compare scope, tailor a resume, draft a note.
The more your AI knows you, the sharper the answers.
Point your assistant at a project or chat that holds your resume, your target titles, the things you will and will not do. jd-intel brings the current openings. Your AI does the matching, the way a sharp friend who knows your work would.
Nothing about your profile leaves your machine to make this work. It is the context your assistant already has.
Built on stable ground.
Public APIs, never scraping
jd-intel reads the same hiring platforms companies publish to. Stable and permitted, where scrapers are fragile and get blocked.
Many platforms, one clean shape
Every posting comes back in one consistent structure your AI can reason over. Plumbing you never have to see.
Runs locally over MCP
A small server on your own machine. Open source, MIT licensed, with nothing hosted in between.
Two minutes. No terminal.
Claude Desktop runs jd-intel for you on its own bundled runtime. Nothing else to install.
Add to Claude Desktop
Download the jd-intel extension, a single .mcpb file.
In Settings, open Extensions, then Advanced settings, and install the file you downloaded.
Start a new chat. The tools appear on their own, and you are ready to ask.
Other clients
Claude Code, Cursor, Windsurf, and VS Code use the same server with one small config.
Setup guide →ChatGPT, Gemini, and other non-MCP clients are not supported yet.