MCP Servers

Extend agent capabilities with Model Context Protocol servers — additional tools and integrations.

MCP (Model Context Protocol) servers extend what agents can do by providing additional tools. They follow an open standard for connecting AI models to external capabilities.

What MCP Servers Do

An MCP server exposes tools that agents can call. For example:

  • A filesystem MCP server might provide tools for listing directories, reading files, and searching content
  • A database MCP server might provide query and schema inspection tools
  • A Slack MCP server might provide messaging and channel management tools

Assigning MCP Servers to Agents

  1. Go to Settings → Agents
  2. Click on the agent
  3. Open the MCP Servers tab
  4. Select which MCP servers this agent should have access to
  5. Changes save automatically

Configuring MCP Servers

To add or manage MCP server configurations:

  1. Go to Settings → MCP Servers
  2. Click Add Server
  3. Configure the server:
FieldDescription
NameDisplay name for this server
CommandThe executable command (e.g., node, python)
ArgsCommand arguments (e.g., path to the MCP server script)
EnvEnvironment variables the server needs
  1. Save the configuration

MCP vs Skills

Both extend agent capabilities, but differently:

SkillsMCP Servers
FormatMarkdown instructions + scriptsRunning process with tool API
How agents use themRead instructions, follow stepsCall tools directly
Best forComplex workflows, multi-step processesDirect tool access, API integrations
SetupContent in databaseServer process configuration