Skip to content

ContextFS

Universal AI Memory Layer — Cross-client, cross-repo context management with semantic search


Features

Feature Description
Semantic Memory Store and retrieve context using semantic search powered by ChromaDB and sentence transformers. Learn more →
Cross-Repo Memories are automatically namespaced by repository, with support for cross-repo search and project grouping. Learn more →
Multi-Client Works with Claude Desktop, Claude Code, Gemini, ChatGPT, and any MCP-compatible client. Learn more →
CLI & MCP Full-featured CLI for memory management plus MCP server for AI tool integration. Learn more →

Installation

pip install contextfs
uv pip install contextfs
pipx install contextfs

Quick Example

from contextfs import ContextFS

# Initialize (auto-detects current repo)
ctx = ContextFS()

# Save a memory
ctx.save(
    content="Authentication uses JWT tokens with 24h expiry",
    type="decision",
    tags=["auth", "security"]
)

# Search memories
results = ctx.search("how does authentication work?")
for r in results:
    print(f"{r.score:.2f}: {r.memory.content}")

CLI Usage

# Save a memory
contextfs save "API uses REST with JSON responses" --type decision --tags api,design

# Search memories
contextfs search "API design patterns"

# Index a repository
contextfs index

# List recent memories
contextfs list

Why ContextFS?

Modern AI development involves multiple tools, repositories, and long-running projects. ContextFS solves the context fragmentation problem:

  • Memory across sessions — Don't repeat yourself to AI tools
  • Memory across tools — Share context between Claude, Gemini, and others
  • Memory across repos — Find related decisions from other projects
  • Semantic search — Natural language queries over your entire context history

Theoretical Foundation

ContextFS is built on principles from Type-Safe Context Engineering, applying insights from protein folding and type theory to AI memory systems.


GitHub PyPI