Overview
The Swarm MCP (Model Context Protocol) server gives AI assistants like Claude and other MCP-compatible tools access to your Swarm data through natural language.
MCP itself does not have a separate credit system. Instead, credit usage is based on the underlying Swarm API endpoints and data that the MCP server accesses on your behalf.
In other words: AI Assistant -> Swarm MCP -> Swarm API -> Credit Consumption
Every MCP action ultimately translates into one or more API requests, and credits are consumed accordingly.
What Counts as a Credit?
Credits are consumed whenever data is retrieved from The Swarm for:
Profile fetches
Company fetches
Intro paths
The exact credit consumption depends on the underlying API endpoints being used.
MCP vs Direct API Usage
Using MCP and using the API directly consume credits from the same credit pool.
For example, these two actions are functionally similar:
Direct API
search_profiles(...)
MCP
Find VP Sales leaders at cybersecurity companies.
Both result in Swarm API requests and consume credits accordingly.
Why MCP Usage Can Vary
With direct API access, developers control exactly which requests are made.
With MCP, the AI assistant determines which tools to call and how many calls are needed to answer a prompt.
For example:
Find everyone at OpenAI who previously worked at HubSpot.
The AI may:
Search for OpenAI
Search for HubSpot
Search matching profiles
Fetch profile details
Refine results
This can result in multiple API calls behind the scenes.
As a result, complex prompts may consume more credits than a single API request.
Example Usage
Low Credit Usage
Prompt:
Who knows Bob Moore?
Typical activity:
Network lookup
Small number of profile lookups
Expected impact:
Low credit consumption
Medium Credit Usage
Prompt:
Find 50 VP Sales leaders at cybersecurity companies.
Typical activity:
Company searches
Profile searches
Profile retrieval
Expected impact:
Moderate credit consumption
High Credit Usage
Prompt:
Research the top 100 AI companies and identify warm introduction paths from our network.
Typical activity:
Hundreds of profile lookups
Company lookups
Relationship mapping requests
Expected impact:
Higher credit consumption
Best Practices
To optimize credit usage:
Consider explicitly capping fetches in your requests
Be specific in your prompts.
Avoid broad research requests unless needed.
Use specifics such as company, title, industry, or geography.
Review MCP activity before running large-scale workflows.
Frequently Asked Questions
Does MCP have its own credit pool?
No. MCP uses the same credits as the Swarm API.
Are credits consumed even when using Claude or ChatGPT?
Yes. The AI assistant accesses Swarm through MCP, which ultimately calls Swarm APIs and consumes credits.
Will MCP consume more credits than the API?
Not necessarily, but it can. AI assistants may perform multiple API calls to answer a single prompt, whereas a direct API integration gives developers complete control over the number of requests being made.
Can I monitor MCP credit usage?
Yes. MCP activity contributes to your overall Swarm credit usage and is reflected in the same credit allocation associated with your workspace.