Skip to content

SpecForgeThe MCP server that gives your AI structured memory

You don't need to be an expert — but experts love it too. Works on Mac, Windows, and Linux. Connects to Claude, Cursor, Windsurf, Gemini CLI, and more. No API key. No account. Install in 2 minutes.

macOS
Windows
Linux

Works with your entire ecosystem

One MCP server. Any AI agent. Any language.

Claude
ChatGPT
Cursor
Windsurf
Gemini
GitHub Copilot
VS Code
Claude
ChatGPT
Cursor
Windsurf
Gemini
GitHub Copilot
VS Code
TypeScript
Python
Go
Rust
Swift
Kotlin
PHP
Ruby
C#
Dart
Flutter
React
Next.js
TypeScript
Python
Go
Rust
Swift
Kotlin
PHP
Ruby
C#
Dart
Flutter
React
Next.js

What problem does it solve?

When building software with an AI assistant, you run into the same problems over and over:

  • The AI forgets what you were building halfway through a conversation
  • You describe the same feature 3 times across different chats
  • The code ends up different from what you originally had in mind
  • You're not sure if the feature is truly "done" or just "kind of working"

SpecForge fixes all of this. It gives your AI a structured memory of what to build, how to build it, and how to verify it's correctly finished — so you stay in control, always.


How it works — 4 steps

1. Tell SpecForge what you want to build

"Create a plan for adding Google login to my project"

SpecForge asks clarifying questions and generates a complete spec — think of it as a short document with a step-by-step plan and a checklist of what "done" looks like.

2. Your AI follows the plan

The spec becomes your development contract. Your AI reads it, follows it, and checks its work against it automatically.

3. Validate when you're done

"Validate spec SPEC-001 against my code at /path/to/project"

SpecForge tells you exactly which items on the checklist are done, which are missing, and what changed from the original plan (we call this "drift" — when the code no longer matches what was agreed).

4. Ship with confidence

Every plan is saved and versioned. Every change is recorded. You always know the real status of your project.


Quick Install

Requires: Node.js ≥ 22 (download here). No API keys. No accounts. All data stays local.

The config block is the same for all agents — only the file location differs:

json
{
  "mcpServers": {
    "specforge": {
      "command": "npx",
      "args": ["-y", "specforge-mcp@latest"]
    }
  }
}

Claude Desktop

1. Open Claude DesktopClaude menu → SettingsDeveloper tab → Edit Config

2. Paste the config block above inside "mcpServers", save, and restart Claude Desktop.

Claude Code — 1 command

bash
claude mcp add specforge -- npx -y specforge-mcp@latest

ChatGPT Desktop

Open or create ~/Library/Application Support/ChatGPT/config.json (Mac) or %APPDATA%\ChatGPT\config.json (Windows), paste the config block, and restart ChatGPT. Requires ChatGPT Plus or Pro.

Cursor

Open or create ~/.cursor/mcp.json, paste the config block, and restart Cursor.

Windsurf

Open or create ~/.codeium/windsurf/mcp_config.json, paste the config block, and restart Windsurf.

Gemini CLI

Open or create ~/.gemini/settings.json, paste the config block, and restart Gemini CLI.

Cline (VS Code)

Open Cline settings → MCP Servers → paste the config block. Cline docs →

Full guide with examples →


Who is it for?

I am...SpecForge helps me...
A solo developer using ClaudeKeep my AI focused and never repeat myself between sessions
A dev new to AI toolsGet structured, professional specs without knowing how to write them
A tech lead or architectEnforce standards, detect drift, and keep the team aligned
A freelancerDocument what I built for clients and detect scope creep early
A student learning to codeLearn professional development practices while using AI tools

See all 59 tools →


59 Tools. 8 Streams.

StreamWhat it does
A — SetupSet up your project so SpecForge understands what you're building
B — PlanningCreate, update, and manage your feature plans (specs)
C — AnalysisEstimate how long things take, audit quality, spot when code drifts from the plan
D — DesignDesign databases, UI contracts, decisions log, step-by-step execution plans
E — Stack & AgentsTool suggestions, version checks, generate custom AI skills
F — DeliveryGenerate tests, documentation, tooling suggestions, connect to project management
G — InfrastructureGit integration, run multiple AI agents in parallel, manage shared context
H — GovernanceAudit dependencies, detect outdated packages, data privacy compliance

See full tools reference →