Agent Setup

Configure TaskFlow with your AI coding agent

Supported Agents

TaskFlow works with any MCP-compatible AI agent. Install the MCP server once, then configure your agent of choice.

Code
bash
1npm install -g @dalmasonto/taskflow-mcp

Quick Comparison

AgentConfig FormatConfig PathAuto-Approve
Claude CodeJSON~/.mcp.json or .mcp.jsonsettings.json allow list
Codex CLITOML~/.codex/config.tomlTop-level approval_policy
CursorJSON~/.cursor/mcp.jsonSettings UI toggle
WindsurfJSON~/.codeium/windsurf/mcp_config.jsonalwaysAllow array
ClineJSONVS Code global storagealwaysAllow array
ContinueYAML/JSON.continue/mcpServers/Agent mode only
GitHub CopilotJSON~/.copilot/mcp-config.jsontools array

How It Works

Every agent communicates with TaskFlow via the Model Context Protocol (MCP) over stdio:

  1. Your agent starts the taskflow-mcp process
  2. The MCP server exposes 50 tools (tasks, projects, timers, analytics, etc.)
  3. The agent calls get_agent_instructions or bootstrap on startup to learn the workflow
  4. From there, the agent autonomously creates tasks, tracks time, and logs progress

No prompt engineering needed - the instructions are built into the server.