Low Context Mode

Low Context Mode is designed for AI apps with limited context windows. Instead of exposing all 50+ tools at once, Macuse provides two meta-tools that let the AI discover and call tools on demand.

When to Use

Enable Low Context Mode when:

Your AI app has a small context window
You're using local LLMs (e.g., via LM Studio, Ollama)
The AI struggles with too many tool definitions

Most cloud-based AI apps (Claude, ChatGPT, etc.) have large context windows and don't need this mode.

How It Works

When Low Context Mode is enabled, Macuse exposes only two meta-tools:

Tool	Purpose
`get_tool_definitions`	Discover available tools and their parameters
`call_tool_by_name`	Execute any tool by name

This reduces the initial context overhead while maintaining access to all functionality.

Two-Step Workflow

The AI follows a discover-then-call pattern:

Step 1: Discover tools

{
  "tool": "get_tool_definitions",
  "arguments": {
    "names": ["calendar_*"]
  }
}

Response includes simplified tool definitions with an example call:

[
  {
    "name": "calendar_create_event",
    "description": "Create a calendar event with title, time, and optional details.",
    "parameters": [
      {
        "name": "title",
        "param_type": "string",
        "required": true,
        "description": "The title of the event"
      },
      {
        "name": "start_date",
        "param_type": "string",
        "required": true,
        "description": "Start date and time in ISO 8601 format"
      },
      {
        "name": "end_date",
        "param_type": "string",
        "required": true,
        "description": "End date and time in ISO 8601 format"
      }
    ],
    "example_call": {
      "tool": "call_tool_by_name",
      "arguments": {
        "name": "calendar_create_event",
        "arguments": {
          "title": "<title>",
          "start_date": "<start_date>",
          "end_date": "<end_date>"
        }
      }
    }
  }
]

Each parameter includes:

Field	Description
`name`	Parameter name
`param_type`	Data type (`string`, `integer`, `boolean`, `array`, `object`)
`required`	Whether the parameter is required
`description`	Optional description of the parameter

The example_call field provides a ready-to-use template for calling the tool.

Step 2: Call the tool

{
  "tool": "call_tool_by_name",
  "arguments": {
    "name": "calendar_create_event",
    "arguments": {
      "title": "Team Meeting",
      "start_date": "2025-01-15T10:00:00",
      "end_date": "2025-01-15T11:00:00"
    }
  }
}

Pattern Matching

The get_tool_definitions tool supports wildcards:

Pattern	Matches
`*`	All tools
`calendar_*`	All calendar tools
`calendar_create_event`	Exact tool name

Enabling Low Context Mode

Per-Connection

Open Macuse Settings
Go to Connections
Find your AI app's connection
Toggle Low Context Mode on

Default for New Connections

To enable Low Context Mode by default for all new connections:

Open Macuse Settings
Go to MCP Server
Enable Default Low Context Mode

New OAuth clients and API keys will have Low Context Mode enabled automatically.

This setting only affects new connections. Existing connections will retain their current Low Context Mode setting. To change the mode for an existing connection, go to Connections and toggle it individually.

Troubleshooting

AI Can't Find Tools

If the AI reports it can't find tools:

Verify Low Context Mode is enabled for the connection
Have the AI call get_tool_definitions with names: ["*"] to list all available tools
Check that the requested toolboxes are enabled in Macuse Settings

Tools Return Errors

If call_tool_by_name returns errors:

Ensure the AI used the exact tool name from get_tool_definitions
Verify all required parameters are provided
Check parameter types match the schema (strings for dates, etc.)

Low Context Mode

On this page