rekal
Active·★ 48·MIT·Updated 2026-05-12
★ Trending★ RAG / Knowledge Base★ Memory & Context
Long-term memory for LLMs. MCP server backed by hybrid search in a single SQLite file.
rekal provides long-term memory for LLMs via an MCP server. It stores memories in a local SQLite file and retrieves them using hybrid search (BM25 keywords + vector semantics + recency decay). It works with any MCP-capable agent like Claude Code, Codex CLI, and OpenCode.
#llm#mcp#mcp-server#mcp-servers#memory#python#sqlite#vector-search
01
Features
01Local SQLite storage, no cloud or API keys
02Hybrid search combining BM25 keywords, vector semantics, and recency decay
03MCP server compatible with various coding agents
04Comprehensive tool-based memory management (store, search, update, delete, prune)
05Smart write operations with supersede and linking for knowledge evolution
02
Compatibility
Claude Code
Claude Code
Verified via docs
Codex CLI
Codex CLI
Verified via docs
OpenCode
OpenCode
Verified via docs
03
Quick start
1
$ pip install rekal
04
Use cases
↳Persistent memory for coding agents across sessions to remember user preferences and decisions
↳Storing project-specific knowledge and conventions without external services
↳Maintaining a searchable history of agent interactions and decisions
05
Alternatives
Brave Search MCP★ 86.5k
Allow your AI Agent to search the real-time internet using Brave Search API. Essential for getting up-to-date information.
FunASR★ 16.6k
Industrial-grade speech recognition toolkit: 170x realtime, 50+ languages, speaker diarization, emotion detection, streaming, and OpenAI-compatible API.
Related searches
Comments
Log in to leave a comment
- SSam ClarkMay 11, 2026
Hybrid search in SQLite for long-term LLM memory is an elegant lightweight solution
- AAlex PatelApr 13, 2026
Used as the memory backend for a coding assistant, recall quality is consistently good
- SSam PatelMar 16, 2026
The hybrid search combines BM25 and semantic retrieval, better than either alone
- TTaylor HarrisMar 6, 2026
Single-file database means zero infrastructure overhead for persistent memory