Check MCP Server

The Check MCP Server connects your AI tools to the Check Payroll API using the Model Context Protocol. Instead of writing integration code for every workflow, use natural language to generate reports, manage employees, run payroll, and more — directly from the AI tools you already use.

Claude Code using the Check MCP to pull information about a company.

🚧
The Check MCP Server is currently in beta. While we've tested it extensively, you may encounter unexpected behavior, breaking changes, or limitations.

No stability guarantees — APIs, tool schemas, and behavior may change without notice between releases.

Not recommended for production-critical workflows — Use at your own discretion and validate outputs before acting on them.

Feedback is welcome — Please report bugs, suggest improvements, or share your experience with your Check point of contact

What you can do

Reports & data exports — Generate payroll journals, tax summaries, employee rosters, and payment reports
Employee & contractor management — Create, update, and view employees and 1099 contractors, manage onboarding and tax forms
Payroll operations — Create payrolls, add earnings and deductions, preview totals, and approve runs
Tax & compliance — View and update W-4 elections, state withholding, tax filings, and exemptions
Company setup & configuration — Manage companies, workplaces, bank accounts, pay schedules, and webhook configurations
Documents & forms — Access tax documents, authorization forms, paystubs, and setup documents
And much, much more

How it works

The Check MCP Server uses a dynamic tool architecture to work within the context window limits of AI models. Instead of loading all 263 tools at once, your AI client starts with three meta-tools:

Tool	What it does
`search_tools`	Find relevant tools by keyword (e.g., "list employees", "create payroll"). Returns matching tools with full parameter schemas.
`run_tool`	Execute any tool by name with the required arguments.
`list_toolsets`	Browse all 17 toolsets with descriptions and example tools.

This means your AI client discovers exactly the tools it needs, when it needs them — keeping context usage efficient while giving you access to the full API surface.

Compatibility

The Check MCP Server works with any MCP-compatible client, including:

Claude Desktop
Claude Code (CLI)
Cursor
GitHub Copilot (VS Code)
ChatGPT
Google Gemini
OpenAI Codex CLI
Any MCP client

Choose your setup

☁️
Hosted (Remote) — Coming soon
Check runs the MCP server for you. Connect directly from your AI client without cloning a repo or managing infrastructure.

🖥️
Self-Hosted — Available now
Clone the repository and run the server locally. Full control over configuration, access control, and environment.

Access control

The MCP server supports fine-grained access control to match your security requirements:

Read-only mode — Disable all write operations (create, update, delete) with a single flag
Toolset filtering — Limit access to specific categories (e.g., only companies and employees)
Tool exclusion — Block individual tools (e.g., exclude create_bank_account)

These options can be combined. See the Self-Hosted guide for configuration details.