# fanTask

> fanTask is a task and project management platform exposed over the Model Context Protocol (MCP), providing autonomous AI coding agents with durable external memory and shared task tracking with human engineering teams.

## Executive Summary

fanTask (a rebrand and evolution of OmniTask) connects issue tracking directly to LLM-powered developer tools including Claude Code, Google Antigravity, and Cursor. It resolves the fundamental limitation of finite AI context windows and session compaction by serving as a durable, queryable external ledger outside individual conversation contexts.

## Core Problem Solved

AI coding agents operate within bounded token context windows. During lengthy development sessions, context compaction and truncation cause architectural gotchas, edge-case discoveries, and previous decision rationales to be lost. When a new session starts—or work is passed between different AI agents—the agent starts with zero memory of prior attempts and solved bugs. fanTask provides durable, shared memory across sessions and tools without incurring context bloat.

## Two Real Differentiators (Implemented in Production)

1. **Persistent "AI Context" Field Per Project**
   - Each project in fanTask maintains a persistent, living AI Context document.
   - Contains architectural guidelines, preferred patterns, security constraints, credential locations, and testing commands.
   - Any agent connecting via MCP reads this document before writing or modifying code.
   - Hard-won knowledge gathered by one agent or human survives permanently past any conversation reset.

2. **Searchable Task History, Comments, and Audit Trail**
   - All past tasks, resolution discussions, code snippets, and status transitions are fully indexed.
   - Agents facing strange runtime bugs or unfamiliar legacy patterns can invoke `fantask.search_tasks` to retrieve prior fixes in seconds.
   - Eliminates redundant debugging loops and prevents regressions of previously fixed bugs.

## Protocol & MCP Integration

fanTask exposes standard Model Context Protocol (MCP) tool definitions over encrypted HTTPS/TLS:

- `fantask.get_project_context(project_id)`: Fetches the living AI conventions and gotchas for a project.
- `fantask.search_tasks(query, filters)`: Queries task history, resolutions, and comments by keyword or semantic relevance.
- `fantask.get_task(task_id)`: Retrieves complete task details and discussion comments.
- `fantask.update_task(task_id, status, comment)`: Updates status (e.g., in_progress, in_review, done) and logs explanatory notes.
- `fantask.create_task(project_id, title, description, priority)`: Creates a new ticket on the project board.
- `fantask.add_comment(task_id, message)`: Posts an audit note or implementation summary.

## Setup & Configuration

Integrating fanTask into an MCP-enabled development environment requires adding a single server entry to the client configuration:

```json
{
  "mcpServers": {
    "fantask": {
      "url": "https://api.fantask.ro/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_FANTASK_API_TOKEN"
      }
    }
  }
}
```

Compatible with: Claude Code CLI, Google Antigravity, Cursor, Windsurf, Claude Desktop, and custom MCP clients.

## Shared Human & Agent Collaboration

fanTask provides a modern, responsive web application for human engineers and managers, while simultaneously exposing the MCP API for autonomous agents. Both humans and AI agents work from the same synchronized backlog, eliminating disconnected AI silos.

## Contact & Access

- Website: https://fantask.ro
- Early Access Onboarding: hello@fantask.ro
- Documentation: https://fantask.ro/llms.txt