ConfigCat MCP Server – Feature Flag Management
ConfigCat's official MCP server exposes its Public Management API to MCP-compatible AI tools. It helps agents inspect and manage feature flags, products, configs, environments, segments, webhooks, audit logs, stale flag reports, and SDK implementation guidance when working in supported clients.
Overview
ConfigCat MCP Server is the official Model Context Protocol server for ConfigCat, a feature flag and configuration management platform. It connects MCP-capable AI clients such as Cursor, Visual Studio Code, and Claude Desktop to ConfigCat's Public Management API so agents can work with live feature flag data rather than relying on static documentation or copied configuration snippets.
What the MCP server enables
The server provides management tools for ConfigCat resources including organizations, members, permission groups, products, configs, environments, segments, SDK keys, webhooks, integrations, code references, audit logs, stale flag reports, tags, feature flags, and setting values. It can create, read, update, replace, and delete many ConfigCat entities, subject to the permissions associated with the supplied Management API credentials. It can also retrieve SDK documentation and code examples to help an editor-based agent integrate ConfigCat SDK usage into a project.
When to use it
Use this MCP server when an AI assistant needs to help maintain ConfigCat configuration from a development environment. Practical workflows include creating a new boolean feature flag, turning a flag on across environments, targeting users in a specific country, cloning an existing flag configuration, updating a flag description, listing stale feature flags, checking audit history for a flag, or inviting a user to a product with a specific permission group.
Connection and authentication
ConfigCat documents a local stdio setup that runs through npx with the package @configcat/mcp-server. The server requires Node.js 16 or later and ConfigCat Public Management API basic authentication credentials. Configure CONFIGCAT_API_USER and CONFIGCAT_API_PASS as environment variables in the MCP client configuration. CONFIGCAT_BASE_URL can be set to override the default Management API host, but this is rarely needed.
Key considerations
The MCP server is intended for management operations, not runtime feature flag evaluation inside production applications. Production applications should evaluate flags through ConfigCat SDKs or ConfigCat Proxy instead. Because the server can modify products, environments, flags, permissions, and webhooks, use least-privilege Management API credentials and require human review for destructive operations. ConfigCat's public API has rate limits; the server respects those limits and returns errors when they are exceeded.
Supported Transports
stdio
Command: npx
Args:
-y@configcat/mcp-server
Frequently Asked Questions
- When should an AI agent use the ConfigCat MCP server?
- Use it when an agent needs live access to ConfigCat feature flag and configuration management from an MCP-compatible client, such as creating flags, updating targeting rules, listing stale flags, reviewing audit logs, or managing ConfigCat products, configs, and environments.
- What does the ConfigCat MCP server add to an AI agent's capabilities?
- It gives the agent tools backed by ConfigCat's Public Management API, enabling the agent to inspect and modify feature flags, settings, tags, segments, webhooks, integrations, members, permissions, products, configs, and environments according to the supplied credentials.
- What can the AI agent access or manage through this server?
- The server documents tools for organizations, members, permission groups, products, configs, environments, segments, SDK keys, webhooks, integrations, code references, audit logs, stale flag reports, feature flags, tags, and setting values, plus SDK documentation for implementation help.
- How is authentication configured for this MCP server?
- Authentication uses ConfigCat Public Management API basic authentication credentials supplied through environment variables. Set `CONFIGCAT_API_USER` for the Management API username and `CONFIGCAT_API_PASS` for the password; keep these values in your MCP client's secret handling rather than hardcoding real credentials in shared files.
- Which transport should be used for this MCP server?
- Use the documented local stdio transport. ConfigCat's setup examples for Cursor, Visual Studio Code, and Claude Desktop run `npx -y @configcat/mcp-server` with the required ConfigCat environment variables; no hosted Streamable HTTP or SSE endpoint is documented for this provider.