Connects Claude to the Open Talent Protocol and Open Job Protocol schemas for parsing and validating structured candidate profiles and job postings. You get six tools: parse raw resumes or job postings into JSON skeletons with extraction placeholders, validate completed documents against the official schemas, and introspect either format into agent-friendly summaries. The parsers are pure functions with no LLM calls or API keys required. They output templates with _EXTRACT_ annotations that your agent fills in, then you validate the result. Useful when you're building recruitment workflows or matching pipelines and need standardized, schema-backed structures instead of freeform text. Runs via npx with zero configuration.
An MCP server for the Open Talent Protocol (OTP) and Open Job Protocol (OJP) — open standards for structuring candidate profiles and job postings for AI agent reasoning.
| Tool | Description |
|---|---|
otp_parse_resume | Raw resume text → OTP skeleton with _EXTRACT_* placeholders for an LLM to fill |
otp_validate_profile | Validate an OTP document against the JSON Schema |
otp_introspect_profile | Extract an agent-friendly summary from an OTP document |
ojp_parse_job_posting | Raw job posting text → OJP skeleton with _EXTRACT_* placeholders |
ojp_validate_job_posting | Validate an OJP document against the JSON Schema |
ojp_introspect_job_posting | Extract an agent-friendly summary from an OJP document |
npm install -g @opentalentprotocol/mcp-server
Or run directly with npx:
npx @opentalentprotocol/mcp-server
Add to your claude_desktop_config.json:
{
"mcpServers": {
"otp-ojp": {
"command": "npx",
"args": ["-y", "@opentalentprotocol/mcp-server"]
}
}
}
Add to your project or global MCP settings:
{
"mcpServers": {
"otp-ojp": {
"command": "npx",
"args": ["-y", "@opentalentprotocol/mcp-server"]
}
}
}
git clone https://github.com/neogene-ai/otp-ojp-mcp-server.git
cd otp-ojp-mcp-server
npm install
npm run build
node dist/index.js
The parsers follow a skeleton + fill pattern:
otp_parse_resume (or ojp_parse_job_posting) with raw text_EXTRACT_* annotations, confidence levels, and known gapsotp_validate_profile (or ojp_validate_job_posting) to verify the resultThis keeps the MCP server LLM-free while giving agents structured guidance on what to extract.
The introspect tools flatten an OTP/OJP document into a normalized, agent-friendly object with an agentSummary string — ready for system prompts, retrieval-augmented context, or matching logic.
The JSON Schemas are bundled in the schema/ directory:
schema/opentalent-protocol.schema.json — OTP v0.1schema/openjob-protocol.schema.json — OJP v0.1MIT