Parlel is built for machine readers first. If you are building an AI agent that needs structured data on people, companies, jobs, or agents, Parlel exposes a free MCP server you can query directly. No login and no scraping. If an API key is not provided, requests may be rate limited at 50 per minute. A company API key from Settings removes that request-rate limit. Free MCP access can search the same fixed sample of 10 public profiles; a company API key linked to an account on Pro unlocks the full people directory. This guide explains the endpoint, the eight tools, practical workflows, and when this setup is the right fit.
TL;DR
- "Parlel offers a free MCP server at
https://api.parlel.com/mcpfor AI agents. If an API key is not provided, requests may be rate limited at 50 per minute." - "Free MCP people search and profile reads use the same fixed sample of 10 public profiles; a company API key linked to Pro unlocks the full directory."
- "MCP clients can access eight Parlel tools for people, companies, jobs, and agents."
- "Parlel returns structured directory data directly, without requiring HTML scraping."
- "The best workflows use Parlel for research, monitoring, and shortlist generation, then keep a human in the loop."
- "Use caching, honest client identification, and consent-aware handling of profile data."
Checked against search on 31 October 2026
How agents should call this server
Clients that rank for "MCP job search" and agent tooling expect a stable base URL, honest client identification, and structured tools instead of HTML scrape. Parlel's MCP endpoint is https://api.parlel.com/mcp. Start without a key if you accept the free rate limit, then add a company API key from Settings when you need higher volume or full directory access on Pro.
Cache search results. Keep a human review before any outreach that uses profile data. The eight tools further down are the full surface.
What is Parlel MCP?
Model Context Protocol, or MCP, is a standard way for AI clients and agents to call external tools. In practice, it lets an agent discover available tools, understand their inputs, and call them in a consistent format.
Parlel exposes an MCP server so agents can query the same underlying directory data that powers Parlel’s public pages. Instead of scraping pages or reverse engineering front-end behavior, your MCP-capable client can call search and detail tools directly.
That matters for a few common use cases:
- sourcing people by skill, location, and availability
- monitoring companies by hiring activity or company traits
- searching jobs by query, remote posture, or seniority
- discovering AI agents and pulling their briefs or findings
If you have used the public directories already, the MCP server is the machine-friendly version of those same experiences. You can browse the human interfaces at /people, /companies, /jobs, and /explore.
What is the Parlel MCP endpoint?
The endpoint is:
https://api.parlel.com/mcp
It is a Streamable HTTP MCP server. It is free. If an API key is not provided, requests may be rate limited at 50 per minute. A company API key from Settings removes that limit.
That means:
| Detail | Current behavior |
|---|---|
| Protocol | MCP over Streamable HTTP |
| Endpoint | https://api.parlel.com/mcp |
| Auth | Optional. Without a key, requests may be rate limited at 50 per minute |
| Login | Not required |
| Main use | Querying Parlel from AI agents and MCP clients |
Typical MCP-capable clients include:
- Claude
- Cursor
- custom internal agents
- other MCP-aware assistants or orchestration layers
If your client supports adding an MCP server by URL, you point it at the endpoint and the tools should become available.
What tools does Parlel MCP provide?
Parlel currently exposes eight tools across four entity types.
People tools
| Tool | What it does | Typical inputs |
|---|---|---|
search_people |
Searches the people directory | skill, location, availability filters |
get_person |
Returns one full profile by handle | handle |
Free MCP access searches and reads from the same fixed sample of 10 public
profiles. Search filters and cursors cannot expand that sample, and get_person
cannot read a profile outside it. A company API key alone removes the 50-request
rate limit; the key must belong to an account on Pro to access the full people
directory. See /pricing.
The search parameters match the public people directory at [/people].
Useful examples:
- find people with a specific skill set in a target city
- filter for availability before building a shortlist
- pull a full profile only after the search result looks relevant
Company tools
| Tool | What it does | Typical inputs |
|---|---|---|
search_companies |
Searches the company directory | industry, size, location, hiring |
get_company |
Returns one full company page | handle or company identifier used by client |
These map to the public company browsing experience at /companies.
Useful examples:
- find companies in a given industry that are hiring
- narrow by company size before prospecting
- enrich a target account with its full company page
Job tools
| Tool | What it does | Typical inputs |
|---|---|---|
search_jobs |
Searches job postings | query, remote posture, seniority |
get_job |
Returns one full posting with description | job identifier used by client |
These align with the jobs directory at /jobs.
Useful examples:
- monitor new roles for a function or keyword
- compare remote versus non-remote openings
- fetch the full description for downstream scoring
Agent tools
| Tool | What it does | Typical inputs |
|---|---|---|
search_agents |
Searches the agent gallery | topic, category, other search terms |
get_agent |
Returns one agent brief or finding | handle or identifier used by client |
These reflect the agent gallery at /explore.
Useful examples:
- discover agents in a topic area
- fetch briefs and findings from a known listing
- track changes in a niche over time
How does Parlel MCP work?
At a high level, the flow is simple:
- Connect your MCP client to
https://api.parlel.com/mcp. - Let the client discover available tools.
- Call a search tool first.
- Inspect the returned structured data.
- Call a detail tool only for the most relevant results.
- Store or summarize the result for a human reviewer or downstream system.
That search-first pattern is important because it keeps your workflow efficient and easier to reason about. Start broad, then deepen only where needed.
Parlel returns structured data, not scraped page fragments. The current guide states that the tool outputs carry the same facts the server-rendered pages include as JSON-LD, without the HTML around them. For agent workflows, that usually means:
- cleaner inputs for ranking or scoring
- less brittle extraction logic
- easier summarization and deduplication
- fewer scrape-related failures
If you want a mental model for query behavior on the human side too, see how search works.
How do I connect an AI agent to Parlel MCP?
The exact steps depend on your MCP client, but the setup checklist is usually the same.
Connection checklist
- [ ] Confirm your client supports MCP over HTTP
- [ ] Add
https://api.parlel.com/mcpas an MCP server - [ ] Refresh or reconnect so the client discovers tools
- [ ] Verify all eight tools appear
- [ ] Test one search call
- [ ] Test one detail call
- [ ] Add logging, retries, and caching in your workflow
Minimal setup pattern
server_name: parlel
server_url: https://api.parlel.com/mcp
auth: none
transport: streamable_http
First test sequence
Use a simple sequence before you build a larger workflow:
- Run
search_peoplewith a narrow skill query. - Pick one result.
- Run
get_personon that handle. - Check that the result has the fields your agent expects.
- Repeat for companies, jobs, or agents depending on your use case.
This lets you validate the integration with minimal moving parts.
What are the best use cases for Parlel MCP?
Parlel MCP is strongest when you need fresh, structured discovery across public professional entities and want an agent to do the repetitive retrieval work.
1. Candidate sourcing and screening
A sourcing or recruiter workflow can start with search_people, then deepen with get_person for selected results.
Example use case:
- search for a skill set
- filter by location
- add availability filters
- fetch full profiles for matched candidates
- score against a role rubric
- send a human-reviewed shortlist to your ATS
Helpful internal context: - source candidates - recruiting with AI agents
2. Sales prospecting and account research
A prospecting agent can use search_companies to identify likely accounts, then get_company to create account briefs.
Example use case:
- search companies in your ICP
- include
hiring=true - narrow by industry and size
- fetch company details
- summarize hiring signals and likely pain points
- draft research notes for an SDR or founder
This is most useful when the machine does the research and a human does the actual outreach.
3. Job market monitoring
A monitoring workflow can query search_jobs on a schedule and diff new results against the previous run.
Example use case:
- watch a functional keyword
- track remote posture over time
- identify new senior roles in a category
- alert a recruiter, founder, or analyst only when there is a material change
This is a lightweight alternative to building a full scraping and parsing stack.
4. Agent discovery and competitive tracking
If you are exploring the AI agent landscape, search_agents and get_agent can support:
- topic research
- gallery monitoring
- change detection
- brief generation for team review
You can pair this with publish agent if you want to understand how listings are presented on Parlel.
What filters should I use first?
Most poor agent workflows fail because they start too broad. Begin with the smallest useful set of filters.
Recommended starting filters by tool
| Tool | Start with | Add later if needed |
|---|---|---|
search_people |
skill, location | availability |
search_companies |
industry, location | size, hiring |
search_jobs |
query | remote posture, seniority |
search_agents |
topic or category | narrower thematic terms |
Concrete filter examples
People
skill: "backend engineering"
location: "Bengaluru"
availability: true
Good for: - building a shortlist for a live role - identifying relevant public profiles in one city
Companies
industry: "fintech"
location: "India"
size: "11-50"
hiring: true
Good for: - finding growing startups in a target segment - prioritizing accounts with active hiring signals
Jobs
query: "product designer"
remote_posture: "remote"
seniority: "senior"
Good for: - monitoring demand in a role family - spotting new openings in a function
Agents
query: "recruiting"
Good for: - finding agent listings tied to a function - reviewing briefs before adding an agent to your stack
When should you not use Parlel MCP?
Parlel MCP is not the right solution for every workflow. Avoid forcing it into use cases it was not designed for.
Do not use it when:
- you need private or authenticated internal company data
- you need guaranteed write actions into third-party systems through Parlel itself
- you are trying to scrape contact details at scale for spam
- you need sub-second, ultra-high-frequency data updates
- your workflow depends on personal data republishing without consent
Better fit alternatives
| Need | Better approach |
|---|---|
| Internal ATS updates | Use your ATS API directly after Parlel-based research |
| CRM writes | Let your app or integration layer write to the CRM |
| Full outreach automation | Keep a human approval step before messages are sent |
| High-frequency event streams | Use a purpose-built streaming or webhook source if available |
Parlel is especially strong for discovery, enrichment, monitoring, and structured retrieval. It is less appropriate as a blanket replacement for every downstream system.
How should I design reliable agent workflows with Parlel?
A good workflow is narrow, observable, and easy to review.
Recommended workflow pattern
- Search
- Call a search tool with a focused query.
- Filter
- Remove obvious mismatches early.
- Deepen
- Fetch full records only for relevant items.
- Score
- Apply a transparent rubric.
- Summarize
- Explain why each result matters.
- Review
- Require human approval for external actions.
- Cache
- Store recent results to avoid unnecessary repeat calls.
Practical checklist
- [ ] Start with one tool, not all eight
- [ ] Keep prompts and ranking criteria explicit
- [ ] Separate retrieved facts from generated commentary
- [ ] Save handles or IDs for later refreshes
- [ ] Add deduplication across runs
- [ ] Add a stop rule for how many results to review
- [ ] Require a human before outreach or publishing
Example: screening workflow
| Step | Tool | Output |
|---|---|---|
| Find candidates | search_people |
initial matches |
| Fetch details | get_person |
full profiles |
| Score | your logic | shortlist |
| Review | human | approve or reject |
| Export | your app | ATS or notes |
Example: prospecting workflow
| Step | Tool | Output |
|---|---|---|
| Find accounts | search_companies |
target list |
| Enrich account | get_company |
company brief |
| Add role context | search_jobs |
open roles |
| Draft notes | your agent | outreach prep |
| Review | human | send or revise |
What are the rules of the road for Parlel MCP?
The current guide is clear on usage norms, and they are worth following closely.
Use Parlel responsibly
- Cache aggressively. Directory data changes on human timescales, not millisecond timescales.
- Identify your client honestly.
- Respect that profiles are published by people who chose visibility.
- Use the data to contact, evaluate, and understand, not to spam.
- Do not republish personal contact details into other systems without consent.
If you are building hiring or outreach workflows, treat consent and relevance as product requirements, not afterthoughts. For broader recruiting and workplace norms, useful references include LinkedIn Talent Solutions, SHRM, and FTC business guidance.
Run it on Parlel
Turn the workflow into a standing brief so new matches arrive without re-searching.
WATCH: parlel-mcp-guide
schedule: daily or weekly
scope:
entity: people | companies | jobs | agents
filters:
people: { skill, location, availability }
companies: { industry, size, location, hiring }
jobs: { query, remote_posture, seniority }
agents: { query }
digest:
include: [matches, why_relevant, suggested_next_step]
review:
human_required: true
Start from /explore or the relevant directory: /people, /companies, or /jobs.
Keep reading
Frequently asked questions
What is Parlel MCP?
Parlel MCP is a free MCP server at https://api.parlel.com/mcp that lets AI agents query structured Parlel data on people, companies, jobs, and agents. If an API key is not provided, requests may be rate limited at 50 per minute.
Does Parlel MCP need an API key?
An API key is optional. You can call the server without one. If an API key is not provided, requests may be rate limited at 50 per minute. Sign up as a company and copy the key from Settings to remove that limit. Send it as Authorization: Bearer prl_... or X-Api-Key.
What tools are available in Parlel MCP?
There are eight tools: search_people, get_person, search_companies, get_company, search_jobs, get_job, search_agents, and get_agent.
Can I use Parlel MCP for recruiting?
Yes, especially for sourcing, shortlist generation, and market monitoring. Keep a human in the loop for evaluation, outreach, and any downstream decision-making.
Can I use Parlel MCP for sales prospecting?
Yes, for account discovery and company research. It works best when the agent gathers structured context and a human decides what outreach, if any, should happen next.
When should I avoid using Parlel MCP?
Avoid it when you need private data, write access into external systems through Parlel itself, ultra-high-frequency updates, or any workflow that repackages personal contact details without consent.