Amnitex: Lossless memory layer for AI coding assistants
atex is a local, lossless memory layer designed for AI coding assistants that supports the Model Context Protocol (MCP) to retain project knowledge across sessions. It offers fast, scalable retrieval using a keyword-scan or spatial tex-grid backend, with sub-microsecond query latency and high recall rates. The tool is open-source, MIT-licensed, and requires no cloud storage, embeddings, or runtime dependencies beyond Python's standard library.
- ▪atex persists project knowledge in a local lossless byte-page key-value store accessible via MCP.
- ▪It supports multiple AI clients like Claude Desktop, Cursor, and Zed by exposing tools such as atex_search and atex_recall.
- ▪The spatial tex-grid backend achieves up to 6075× speedup over keyword scanning while maintaining 95–100% recall@1.
- ▪atex has been validated end-to-end using a 0.5B open-source model via Ollama with exact recall of stored facts.
- ▪Storage is structured in raw byte pages and a JSON index, ensuring transparency and avoiding proprietary formats.
Opening excerpt (first ~120 words) tap to expand
atex (amni-tex) A lossless byte-page memory layer for MCP-capable AI coding assistants. Local. No embeddings. No cloud. MIT-licensed. Zero runtime dependencies beyond stdlib. Install name on PyPI is amnitex; the command-line tool itself is atex. pipx install amnitex cd /path/to/your/project atex init atex demo # auto-detects MCP clients (Claude Desktop, Claude Code, Cursor, Cline, Continue, Zed) and wires the config with [y/N] consent That's it. Restart your AI client; atex_search, atex_recall, atex_remember, atex_list_keys, atex_stats are now available as tools. What it does Every AI coding assistant forgets your project the moment a new session starts.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at Hacker News: Newest.
Discussion
0 commentsMore from Hacker News: Newest
