AI candidate scoring via MCP. Score resumes against job descriptions with Claude — returns dimension scores, strengths, and gaps.

What is AI candidate scoring via MCP?

AI candidate scoring via MCP is a source-available MCP server for HR and recruiting workflows. It scores candidate resumes against job descriptions using Claude, returning structured dimension scores, strengths, and gaps. Any agent framework that speaks MCP can call it natively.

How to use AI candidate scoring via MCP?

Use the hosted API at recruitapi.app without running anything, or self-host by deploying to Railway in under 15 minutes. Send a POST /score-candidate request with resume_text and job_description fields, and optionally custom weights and recency_window_years. Pass an API key via X-API-Key or Authorization: Bearer header.

Key features of AI candidate scoring via MCP

• Scores candidates on four configurable dimensions with overall score (0–100)
• Returns 2–4 grounded strengths and gaps per candidate
• Supports idempotency caching for 24 hours with a header
• Self-hostable on Railway or locally with Anthropic API key
• Credit-based billing: 100 credits for $5, expire after 180 days
• Rolling rate limits: 30 requests/minute and 500 requests/day per key

Use cases of AI candidate scoring via MCP

• Automatically score a batch of resumes against a job description
• Integrate candidate scoring into an existing agent or workflow via MCP
• Self-host a private scoring system without sending data to third parties

FAQ from AI candidate scoring via MCP

How do I get an API key for AI candidate scoring via MCP?

API keys are issued after a credit purchase. Visit hrmcp-server-production.up.railway.app/billing to buy credits and receive a key.

What are the input requirements for the resume and job description?

Both text fields must be between 50 and 15,000 characters. The resume must be at least 50 words; the job description may trigger a warning if under 30 words.

What happens if I exceed the rate limits?

The server returns a 429 status with code rate_limit_exceeded. Each response includes X-RateLimit-* headers showing remaining capacity. Self-hosted deployments can disable rate limiting with RATE_LIMIT_ENABLED=false.

Can I run AI candidate scoring via MCP locally without a database?

Yes. Scoring works without a database. Auth and billing require Postgres, but local development can test scoring with just ANTHROPIC_API_KEY.

What weights does AI candidate scoring via MCP use by default?

Default weights are: skills_match 0.40, experience 0.30, industry_background 0.20, education 0.10. Custom weights can be provided and must sum to 1.0 with each value between 0.0 and 1.0.