ContextForge

About ContextForge

ContextForge gives AI coding assistants a memory that survives between sessions, exposed through the Model Context Protocol so the same store is shared across every tool that speaks it. The problem it addresses is mundane and expensive: you open Claude Code, re-explain the architecture, re-explain the naming conventions, re-explain the auth decision made last week, and pay for those tokens every single time. Setup is one JSON block added to an MCP config pointing at an npx-run server with an API key, after which the assistant gains tools for reading and writing persistent memory. Supported clients include Claude Code, Claude Desktop, ChatGPT desktop, Cursor, Windsurf and GitHub Copilot in VS Code agent mode. On top of raw storage it adds semantic search across the knowledge base, git integration that auto-syncs commits and pull requests so the memory reflects what actually happened in the repository, snapshots for backup and restore, import and export, webhooks for change notifications, and an organisational model of projects containing spaces so that context stays scoped to the right repository rather than bleeding across unrelated work. Collaborators can be invited to a project on the paid tiers. The free tier is real rather than a teaser — 500 semantic queries a month, 200 documents, three spaces — which makes it viable for a single personal project.

Key Features

Tags