MCP Catalogs
Home

forge-orchestrator

by nxtg-ai·116·Score 47

Rust-based MCP server for multi-AI task orchestration with file locking, knowledge capture and drift detection.

developer-toolsai-llmproductivity
17
Forks
3
Open issues
this month
Last commit
2d ago
Indexed

Overview

Forge Orchestrator is a sophisticated MCP server designed to coordinate multiple AI coding tools on shared repositories. It solves the problem of multiple AI tools working on the same codebase without shared state, providing file locking to prevent conflicts, knowledge capture to maintain decision context across sessions, and drift detection to ensure work stays aligned with specifications. The server exposes 11 tools through MCP stdio protocol, making it compatible with any connected AI client. Built in Rust as a single binary with no runtime dependencies, it's designed for both interactive dashboard use and headless autonomous execution.

Try asking AI

After installing, here are 6 things you can ask your AI assistant:

you:Coordinating multiple AI coding tools (Claude Code, Codex CLI, Gemini CLI) on the same repository
you:Preventing file conflicts through exclusive locking mechanism when multiple AI tools work on shared code
you:Capturing and reusing knowledge across different AI tool sessions to maintain context
you:What AI tools does Forge Orchestrator support?
you:How does file locking work?
you:What is the knowledge flywheel?

When to choose this

Teams using multiple AI coding tools on the same repository who need coordination, file locking, and knowledge preservation between tool sessions.

When NOT to choose this

If you're only using a single AI coding tool, or need a fully open-source solution (this uses a functional source license that converts to Apache 2.0 in 2028).

Tools this server exposes

11 tools extracted from the README
  • init

    Initialize Forge in a project with state, event log, and knowledge base

  • plan_generate

    Generate task plan from specifications by decomposing into dependency-aware tasks

  • dashboard

    Launch the TUI dashboard with live tool panes and task status

  • run

    Execute tasks headlessly in autonomous mode for CI/CD pipelines

  • status

    Show current state summary including task assignments and dependencies

  • verify

    Run automated acceptance tests on the project

  • uat

    Run interactive user acceptance tests

  • ship

    Release ceremony with changelog generation, archiving, and tagging

  • config_brain

    Configure the AI brain with openai API key or rule-based approach

  • drift_detection

    Compare in-progress work against specifications to detect divergence

  • file_locking

    Acquire or release exclusive file locks to prevent conflicting edits

Note: Tool names inferred from CLI commands in the README. While it mentions there are 11 MCP tools, the exact MCP API signatures and tool names aren't explicitly documented, only the CLI commands. Tools like 'drift_detection' and 'file_locking'

Comparable tools

agentic-orchestratormulti-agent-toolkitai-workflow-platform

Installation

Install

curl -fsSL https://forge.nxtg.ai/install.sh | sh
forge init
Claude Desktop Integration

Add to Claude Desktop config:

{
  "mcpServers": {
    "forge": {
      "command": "forge",
      "args": ["mcp"]
    }
  }
}
From Source
git clone https://github.com/nxtg-ai/forge-orchestrator
cargo build --release

Note: Single binary, 4.7 MB. No runtime dependencies.

FAQ

What AI tools does Forge Orchestrator support?
It supports Claude Code (via MCP stdio), Codex CLI, and Gemini CLI (via filesystem). Each tool adapter reads its own config format without requiring a specific configuration language.
How does file locking work?
When a tool starts editing a file, Forge acquires an exclusive lock in .forge/locks/. Other tools requesting the same file are queued and notified. The system includes timeouts for crashed tools and deadlock detection, tested with 378 test cases covering concurrent access scenarios.
What is the knowledge flywheel?
The knowledge flywheel captures decisions, patterns, and learnings from all tool sessions in .forge/knowledge/. Entries are auto-classified and searchable across tools and sessions. This context is preserved for future use, allowing new sessions to start with accumulated knowledge rather than from scratch.

On Hacker News

Recent discussion from the developer community.

Compare forge-orchestrator with

GitHub →

Last updated · Auto-generated from public README + GitHub signals.