timrogers/suggest-follow-up-task-mcp-server

suggest-follow-up-task-mcp-server

A basic MCP (Model Context Protocol) server to allow AI agents to propose follow-up tasks.

The follow-up tasks proposed don't get stored anywhere - but adding a tool like this is a way to elicit suggestions from the model, and record them to the agent logs for future analysis.

Usage with GitHub Copilot coding agent

1. Update your repository's MCP configuration, adding the server

{ 
  "mcpServers": {
    "suggest-follow-up-task": {
      "type": "local",
      "tools": ["*"],
      "command": "npx",
      "args": ["-y", "github:timrogers/suggest-follow-up-task-mcp-server"]
    }
  }
}

2. Add a .github/copilot-instructions.md file to prompt the model to use the tool (optional)

As you work, watch out for potential follow-up tasks in this repository that are likely to be valuable or useful.

When you spot such a task, call the suggest-follow-up-task-propose_task with a short description.

When you finish work, think deeply about this again, and propose any further follow-up tasks that you've missed.

3. Ask Copilot to work on a task. In the logs, you will see follow-up tasks proposed by Copilot.

Features

• propose_task tool: Records a task and returns confirmation
• Minimal, focused implementation
• TypeScript-based with full type safety
• Stdio transport for MCP communication

Available tools

propose_task

Suggest a follow-up task to complete in the future in this codebase. As you work, if you identify follow-up tasks that are likely to be valuable or useful, call this tool to propose them. When you finish work, consider again, and call this tool to propose any valuable follow-up tasks you haven't yet proposed.

Arguments:

• task (string, required): The follow-up task to propose as an imperative statement (e.g., "Add unit tests for transformWidget function in src/widgets.ts" or "Handle empty string in title validation in app/models/ticket.rb"). Avoid vague phrases like "Consider..." or "Think about..."

Returns: "Task recorded" message

Local development

# Install dependencies
npm install

# Build TypeScript
npm run build

# Run in development mode
npm run dev

# Run tests
npm test

# Run tests with coverage
npm run test:coverage

# Run tests in watch mode
npm run test:watch

# Run all checks (lint, format, build, test)
npm run check

# Fix linting and formatting issues
npm run fix

Testing

The project includes comprehensive integration tests that verify:

• Server startup and shutdown
• MCP protocol communication via stdio
• Tool listing and execution
• Error handling
• JSON-RPC message handling

Run tests with:

npm test

CI/CD and Code Quality

This project uses GitHub Actions for continuous integration with the following features:

• Multi-version Node.js testing: Tests run on Node.js 18.x and 20.x
• Comprehensive checks: Linting (ESLint), formatting (Prettier), building (TypeScript), and testing (Jest)
• Reviewdog integration: Automated PR review comments for linting and TypeScript compilation issues
  • ESLint issues are posted as inline PR comments
  • TypeScript compilation errors are highlighted directly in the PR
  • Only runs on pull requests for focused feedback

The CI workflow ensures code quality while providing developers with immediate, actionable feedback during the review process.