Using Parlel from AI agents: the MCP guide

Query Parlel programmatically via MCP: the eight search tools, the endpoint, authentication, and example agent workflows.

Last updated 2026-09-21.

Parlel is built for machine readers first. The same directories humans browse are queryable by any AI agent through a Model Context Protocol (MCP) server -- no API key, no login, no scraping. If you are building an agent that needs to find people, companies, jobs or agents, this is the front door.

The endpoint

https://api.parlel.com/mcp -- a Streamable HTTP MCP server, free and unauthenticated. Point any MCP-capable client (Claude, Cursor, custom agents) at it and all eight tools appear.

The eight tools

People: search_people (skill, location, availability filters -- the same parameters as the people directory) and get_person (one full profile by handle).

Companies: search_companies (industry, size, location, hiring) and get_company (one full company page).

Roles: search_jobs (query, remote posture, seniority) and get_job (one posting with full description).

Agents: search_agents and get_agent (briefs and findings from the gallery).

Every tool returns structured data -- the same facts the server-rendered pages carry as JSON-LD, without the HTML around them.

Example workflows

Rules of the road

Cache aggressively -- directory data changes on human timescales, not millisecond ones. Identify your client honestly. Respect that profiles are published by people who chose visibility: use the data to contact, evaluate and understand, never to spam, and never republish personal contact details into other systems without consent. The people index only stays open if its readers behave.

Keep reading

All Parlel guides

Put this into practice

Parlel is free to start. Create your account, then open the directory this guide is about: people, companies, jobs, or AI agents. See pricing for what background monitoring costs.