Toolselect

You have an MCP server that suggests which tools to use for a given task, learning from usage patterns to improve recommendations over time. It helps you quickly pick effective tools, tracks outcomes, and tunes suggestions as you work, saving time and improving accuracy across tasks.

How to use

To use this MCP server, connect your MCP client with the toolselect endpoint configured as shown in the configuration examples. You can describe your task to receive ranked tool recommendations, each with a confidence score, a priority level, and historic success data. After you run a recommended tool, record the outcome so future recommendations become more accurate. The system analyzes the task description to extract keywords and estimate complexity, then matches those signals to registered tools based on strengths and historical results. You’ll receive prioritized recommendations that help you choose the most suitable tool for your current work.

How to install

Prerequisites: you need a runtime that can execute MCP tooling. This setup uses Bun as the package manager and runner.

# Install the MCP toolselect package
bun install @aegis-ai/mcp-toolselect

# Or clone the repository and run directly
git clone https://github.com/aegis-ai/mcp-toolselect.git
cd mcp-toolselect
bun install
bun src/index.ts

Configuration

Configure your MCP client to connect to the toolselect server. You can run the server locally or use a remote instance. The following examples show two common configurations that you can paste into your MCP client configuration file.

{
  "mcpServers": {
    "toolselect": {
      "command": "bunx",
      "args": ["@aegis-ai/mcp-toolselect"]
    }
  }
}

If you are running from source locally, you can configure the client to start the module directly using the path to the source index. Use this approach when you are actively developing or testing changes.

{
  "mcpServers": {
    "toolselect": {
      "command": "bun",
      "args": ["/path/to/mcp-toolselect/src/index.ts"]
    }
  }
}

Practical notes about usage

• The server exposes a function to get ranked tool recommendations for a task description. You can specify how many results you want, and you’ll receive details such as confidence, priority, and historical success rates.

Tools and capabilities overview

The server supports a set of core tools that enable searching, registering, and tracking the performance of tools within your workflow.

Available tools

recommend_tools

Get ranked tool recommendations for a task description, including confidence scores, priority levels, and historical success rates.

register_tool

Register a tool with its capabilities so it can be recommended for future tasks, including name, description, category, strengths, and use cases.

record_usage

Record that a tool was used for a task and whether it succeeded, providing feedback to improve future recommendations.

get_tool_stats

Retrieve usage statistics and success rates for registered tools.

list_tools

List all registered tools grouped by category.