Universal AI Memory Layer for developers. Persistent, searchable context that follows you across sessions, repos, and AI tools.
Your coding decisions, bug fixes, and architectural knowledge persist across sessions. Never re-explain context to your AI assistant.
ChromaDB + sentence-transformers for intelligent retrieval. Find relevant context by meaning, not just keywords.
Memory isolation per repository with optional sharing. Team knowledge stays organized and accessible.
Smart routing: FTS for session/keyword search, RAG for semantic code search. Best of both worlds.
Works with Claude Code, Claude Desktop, Gemini CLI, Codex CLI, and any MCP-compatible client.
Standard Model Context Protocol server for seamless integration with AI coding assistants.
Capture architectural decisions, library choices, and design patterns as you work.
Record error patterns and their solutions. Never debug the same issue twice.
Save deployment steps, setup guides, and common workflows for instant recall.
Semantic search finds relevant context. Your AI assistant gets full project history.
# Save a decision ctx.save( "Use PostgreSQL for the database", type=MemoryType.DECISION, tags=["database", "architecture"] ) # Record a bug fix ctx.save( "CORS fix: allow_methods=['*']", type=MemoryType.ERROR, tags=["cors", "api", "bug-fix"] ) # Search for context results = ctx.search("database decisions") # Smart routing automatically uses: # - FTS for sessions/keywords # - RAG for code/semantic
Team members share a common namespace for project-wide knowledge.
New team members instantly access all architectural decisions and conventions.
Link related repos (frontend/backend) with shared context about integrations.
PostgreSQL sync for team-wide memory sharing. Local SQLite for speed.
# Team namespace ctx = ContextFS(namespace_id="team-acme") # Save team conventions ctx.save( "API uses camelCase for JSON keys", type=MemoryType.FACT, tags=["api", "convention", "team"] ) # Cross-repo context ctx.save( "Frontend calls /api/v2/analyze", type=MemoryType.FACT, tags=["integration", "api"] ) # New team member searches results = ctx.search("API conventions") # Gets all team knowledge instantly
Project facts, configurations, conventions
Architectural decisions with rationale
Important algorithms, patterns, snippets
Bug fixes, error patterns, solutions
How-to guides, deployment steps
Session transcripts, conversations
+------------------+
| AI Assistants |
| Claude, Gemini, |
| Codex, Custom |
+--------+---------+
|
MCP Protocol
|
+--------v---------+
| ContextFS |
| MCP Server |
+--------+---------+
|
+------------------+------------------+
| | |
+-------v-------+ +-------v-------+ +------v-------+
| SQLite + | | ChromaDB | | Session |
| FTS5 | | (Vector) | | Manager |
+---------------+ +---------------+ +--------------+
| Keyword Search| |Semantic Search| | Capture & |
| Fast Queries | | Code Similar. | | Replay |
+---------------+ +---------------+ +--------------+
| | |
+------------------+------------------+
|
+--------v---------+
| Namespace |
| Isolation |
| (per-repo/team) |
+------------------+