OpenCaddis Now Supports MCP Servers — Connect Your Agents to Any External Tool

Eric Brasher February 21, 2026 at 7:35 AM 5 min read

OpenCaddis agents can now connect to MCP (Model Context Protocol) servers, giving them access to tools from services like GitHub, Slack, databases, file systems, and hundreds more — all configured through the UI with no code changes.

What is MCP?

MCP is an open standard for connecting AI agents to external tools and data sources through a standardized protocol. Rather than building a custom plugin for every integration, an agent connects to an MCP server and automatically discovers all the tools it provides. One protocol, unlimited integrations.

How It Works

When you add an MCP server to an agent in OpenCaddis, the underlying FabrCore runtime handles the connection automatically during agent activation. The agent calls ResolveConfiguredToolsAsync(), which connects to each configured MCP server, discovers its tools, and registers them alongside the agent's built-in plugins and tools. The agent doesn't need any awareness of MCP — it just sees more tools available to use.

Two transport types are supported:

  • Stdio — Launches a local process on the host machine. The agent communicates with it over stdin/stdout. This is the most common type, used by servers like @modelcontextprotocol/server-github, @modelcontextprotocol/server-filesystem, and most community MCP servers.
  • Http — Connects to a remote HTTP endpoint. Used for hosted or shared MCP servers accessible over the network.

Configuration

MCP servers are configured per-agent through the Settings UI under Agent Configuration. Edit any agent, scroll to the MCP Servers section, and add as many servers as you need. Each agent can connect to multiple MCP servers simultaneously — all discovered tools are merged together.

For Stdio servers, you specify the command, arguments, and environment variables. For Http servers, you specify the URL and any headers needed for authentication.

The configuration is stored in opencaddis.json and supports export/import, so you can share agent configurations across environments.

Example: GitHub

Add a single MCP server to any agent to give it full GitHub access:

Namegithub
TransportStdio
Commandnpx (or full path on Windows: C:\Program Files\nodejs\npx.cmd)
Arguments-y, @modelcontextprotocol/server-github
Environment VariablesGITHUB_PERSONAL_ACCESS_TOKEN = your PAT

Save and reload. The agent now has tools for creating issues, searching repositories, reading files, managing pull requests, listing commits, and more — all discovered automatically from the MCP server.

What's Next

The MCP ecosystem is growing fast. Servers exist for Slack, PostgreSQL, SQLite, Google Drive, Brave Search, Puppeteer, and many others. Any MCP-compatible server works with OpenCaddis out of the box.

Check out the full MCP documentation for transport type details, example configurations, the config file format, and troubleshooting tips.


Eric Brasher

Builder of OpenCaddis and the FabrCore framework.