FoundRole MCP Server — AI Job Search in ChatGPT, Claude & Cursor

Ask ChatGPT or Claude for jobs and get real openings back — each one checked for whether it's still real, what it actually pays, and whether the company sponsors visas. The FoundRole MCP server connects your AI assistant to FoundRole's live job board, application tracker, and career knowledge base, so the whole job search runs inside the chat you already use.

Sign in once — your assistant handles it. The first time your AI calls FoundRole it opens a standard OAuth sign-in; approve it once in the browser and that covers everything: search, fact-checks, the tracker, reminders, and alerts. There is no API key to copy or rotate, and the account is free.

What your assistant can do

• Search live jobs by asking. Openings straight from company career pages, refreshed hourly across 40+ industries. Natural language maps onto real filters — job title, location, company, salary range, posting date: "remote React jobs in NYC paying over $130k posted this week."

• Get every posting fact-checked — free. Before you spend an evening on an application, three questions get answered with the evidence behind each call:

  • Is anyone actually hiring? Ghost-posting risk — whether the role has been reposted for months or looks like it's fishing for resumes.

  • Am I being underpaid? The pay against what employers in that market really file, so a hidden salary range stops being a guess.

  • Would they sponsor me? Whether the company has sponsored visas before, counted year by year from public filings.

• See how well you fit. Match scoring against your resume and profile, plus personalized job recommendations ranked by fit.

• Research before you commit. Dig into companies, industry sectors, and hiring by location to decide where to aim.

• Compare roles side by side. "Compare these two" lines up fit, pay, sponsorship, and posting risk at once — so the choice stops being a feeling.

• Paste a job from anywhere. LinkedIn, a careers page, a link a friend sent — paste the posting into the chat and it gets the same checks, and can sit in your tracker alongside FoundRole listings. Your assistant reads the text you provide; nothing crawls the site for you.

• Stop losing track of applications. Save jobs to your Kanban application tracker and move them through Saved → Applied → Interviewing → Offered → Hired by asking. Attach notes, tags, expected salary, recruiter contacts, and deadlines. The same board shows up in the web app.

• Never miss a follow-up. Set a reminder on any tracked job and get an email with a calendar (.ics) invite for Google Calendar, Outlook, or Apple Calendar.

• Let the search come to you. Subscribe a search to recurring email alerts — daily, weekly, or monthly — so new matches land in your inbox.

• Ask the awkward questions too. Interview prep, salary negotiation, resume tactics, company research — answers grounded in FoundRole's knowledge base, with the sources linked.

• See results as real panels, not text walls. In clients that support MCP Apps (ChatGPT among them), search results, job details, and your tracker render as interactive panels right in the chat.

It's the same account and data as FoundRole.com — the MCP server just lets your AI drive it.

Related MCP server: jd-intel-mcp

Free vs Pro

Search, the fact-checks on every posting, the tracker, reminders, and alerts are free with no usage limits. Pro adds screening: ask for remote-only, sponsors-only, a salary floor, risky postings hidden, or a minimum match, and your assistant filters the list before you ever see it — and tells you when it has. A free account still gets the full list back, with a note that those filters were not applied.

New here? The FoundRole AI Search guide walks through connecting each client step by step, with an FAQ.

Setup

Two ways to connect, depending on your client. Either way, your client opens a FoundRole sign-in the first time it connects — approve it once and you're set.

Option 1 — Remote server (recommended)

Point your client at the FoundRole MCP endpoint:

https://www.foundrole.com/mcp

Most modern clients — ChatGPT, Claude, Cursor, VS Code — speak remote MCP (Streamable HTTP) natively, so this is all the configuration there is.

Option 2 — stdio bridge (for clients without remote MCP)

If your client only supports stdio transport, run this package locally with npx; it bridges stdio to the FoundRole endpoint and handles the OAuth sign-in for you:

npx @foundrole/ai-job-search-mcp

Connecting your AI assistant

ChatGPT

Estimated time: ~1 minute

1. Open FoundRole in the ChatGPT app directory.

2. Click Add and approve the FoundRole sign-in when ChatGPT opens it.

3. Ask for jobs in any chat — no settings to configure, nothing to paste.

Claude Web/Desktop

Estimated time: ~2 minutes

1. Open Claude settings (profile / settings icon).

2. Find Connectors (or Tools) and click Add custom connector.

3. Name it FoundRole.

4. In the Remote MCP server URL field, paste:

   https://www.foundrole.com/mcp

5. Save, allow Claude to connect, and approve the FoundRole sign-in it opens. Then ask for jobs in the chat.

Cursor / VS Code / Windsurf

In Cursor, open Settings → MCP and add this to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "foundrole": {
      "url": "https://www.foundrole.com/mcp"
    }
  }
}

In VS Code, add this to .vscode/mcp.json in your workspace (or your user mcp.json) — VS Code uses a different key and needs an explicit transport type:

{
  "servers": {
    "foundrole": {
      "type": "http",
      "url": "https://www.foundrole.com/mcp"
    }
  }
}

Approve the FoundRole sign-in when your editor opens it, then ask for jobs in the chat.

LM Studio

Estimated time: ~3 minutes | Requires: Node.js + npm

1. Open LM Studio's MCP servers configuration.

2. Add this entry:

   {
     "mcpServers": {
       "foundrole": {
         "command": "npx",
         "args": ["@foundrole/ai-job-search-mcp@latest"],
         "env": {}
       }
     }
   }

3. Save and restart LM Studio.

4. Confirm foundrole appears in your MCP servers list, approve the sign-in it opens, and start asking for jobs.

Perplexity Desktop

Estimated time: ~4 minutes | Requires: Perplexity Pro + Node.js

1. Open Perplexity's MCP configuration.

2. Add:

   {
     "command": "npx",
     "args": ["@foundrole/ai-job-search-mcp@latest"],
     "env": {}
   }

3. Save and restart Perplexity, then approve the FoundRole sign-in when it opens.

Perplexity uses stdio MCP transport — Node.js and npm must be installed.

Any other MCP client

The same address works everywhere: add https://www.foundrole.com/mcp as a remote server, or bridge stdio-only clients with npx @foundrole/ai-job-search-mcp.

Note: After you approve the sign-in, some clients need a restart (quit and reopen) before the connection goes live.

Try saying

• "Find senior backend engineer roles in San Francisco posted this week."

• "Which of these is least likely to be a ghost posting?"

• "Does Stripe sponsor visas? What does this role really pay?"

• "Compare the Stripe and Vercel roles on fit, pay, and sponsorship."

• "Here's a posting from LinkedIn — run the same checks on it."

• "Recommend jobs that match my resume."

• "Save it to my tracker and mark it Applied."

• "Remind me to follow up next Tuesday morning."

• "What's in my Interviewing column right now?"

• "Subscribe me to weekly alerts for remote React jobs."

• "How do I answer 'walk me through your resume'?"

Security

• OAuth 2.1 with PKCE. You sign in to FoundRole through a standard authorization flow — short-lived tokens, instant revocation, no credentials shared with the AI client, no API key to copy, store, or leak.

• HTTPS only, with dynamic client registration and redirect-URI validation.

• Streamable HTTP transport (direct), or stdio via this proxy.

• It never applies or emails anyone as you.

Troubleshooting

Asked to sign in / "needs authentication":

• Expected on first connect — the server authenticates every session. Complete the FoundRole sign-in in the browser window your client opens.

• If tools still don't appear afterward, restart the client or reconnect the connector so the authorization is re-sent.

"Connection failed":

• Check your internet connection and that the URL is exactly https://www.foundrole.com/mcp.

• Confirm your client supports remote HTTP MCP; if not, use the stdio bridge (Option 2).

"Command not found" (stdio clients):

• Install Node.js (see engines in package.json for the required version), then retry, or install globally: npm install -g @foundrole/ai-job-search-mcp and run ai-job-search-mcp.

Connector not working:

• Double-check the URL, complete the sign-in, and restart the AI client — some clients only pick up the connection after a restart.

Explore the data behind the answers

The same data your assistant reads is browsable on FoundRole:

• How It Works — where the openings come from and how each posting gets its ghost, pay, and visa checks

• H1B Salary Explorer — certified wages employers filed with the U.S. Department of Labor, charted by sector, industry, location, and role

• H1B Sponsor Rankings — top visa-sponsoring companies by median filed wage, browsable by industry, sector, city, and state

• Company Directory — open roles, salary data, and visa sponsorship history for any employer

• Industry Sectors — who's hiring across 40+ industries

• Hiring by Location — openings and wage benchmarks by state and metro

• Job Tracker — the Kanban board your assistant manages, in the browser

• Live Job Board — browse the listings directly

Help

• 🌐 Website: www.foundrole.com — free AI job search, application tracker, and company research

• 📖 Setup guide & FAQ: foundrole.com/ai-search-mcp

• 🐛 Issues: GitHub Issues

• 💬 Questions: dev@foundrole.com

Connect your AI client, sign in once, and run the whole job search from the chat — search, fact-checks, tracking, reminders, and alerts, free at FoundRole.

Made by the FoundRole team.