Contextforge Mcp
Give Claude Code, Cursor, or Copilot durable project memory through ContextForge instead of losing context every new session.
Overview
ContextForge MCP is a MCP server for the Build phase that adds hosted persistent memory for Claude Code, Cursor, and GitHub Copilot sessions.
What is this MCP server?
- Persistent memory MCP for Claude Code, Cursor, and GitHub Copilot (contextforge-mcp v0.2.0)
- Requires CONTEXTFORGE_API_KEY from contextforge.dev dashboard
- Optional CONTEXTFORGE_API_URL override (default https://api.contextforge.dev)
- stdio npm transport (identifier contextforge-mcp)
- Server package version 0.2.0 (contextforge-mcp on npm)
- stdio transport
- 1 required secret: CONTEXTFORGE_API_KEY; optional CONTEXTFORGE_API_URL defaulting to https://api.contextforge.dev
What problem does it solve?
Long build cycles stall when every new agent chat forgets stack choices, naming rules, and half-finished plans you already documented once.
Who is it for?
Solo builders on multi-week repos who switch between Claude Code, Cursor, or Copilot and need one shared memory substrate.
Skip if: One-off scripts, teams forbidding external memory SaaS, or builders who only need ephemeral in-chat context with no cross-session recall.
What do I get? / Deliverables
After you configure contextforge-mcp with an API key, agents can read and write ContextForge memory so context carries across sessions and tools.
- MCP tools to store and recall project memory in ContextForge
- Shared memory usable across supported coding agents
- Configurable API endpoint via CONTEXTFORGE_API_URL
Recommended MCP Servers
Journey fit
Persistent agent memory is a build-time multiplier for solo developers who iterate for weeks in the same repo with multiple tools. Agent-tooling captures MCP-backed memory that rides alongside coding sessions rather than shipping, SEO, or production incident response.
How it compares
Hosted persistent-memory MCP, not encrypted multi-agent chat or cloud cost FinOps.
Common Questions / FAQ
Who is ContextForge MCP for?
Indie developers using Claude Code, Cursor, or GitHub Copilot who want cross-session memory without manually pasting project briefs each time.
When should I use ContextForge MCP?
Use it at the start of a build engagement and throughout implementation whenever you need agents to remember architecture, preferences, and prior decisions.
How do I add ContextForge MCP to my agent?
Sign up at contextforge.dev, copy CONTEXTFORGE_API_KEY into your MCP client env, install contextforge-mcp (v0.2.0) via stdio, optionally set CONTEXTFORGE_API_URL, then enable the server in Claude Code or Cursor.