ais — Claude Code MCP server: what it does, how to install
Plain-text local memory for AI agents. It maps user-defined keywords to notes and file paths, returning exact references without vector databases.
Open on GitHubWhat ais does
ais maintains an associative catalog directly in plain text on your drive. Instead of running semantic approximations over embeddings, the tool indexes records strictly under the words you assign. Querying multiple labels returns items associated with all of them simultaneously. Target files or remote links stay where they are, as the index stores only pointers without duplicating data.
The core engine is compiled C without external runtimes or database engines. A command-line program serves as the primary layer. An agent accesses ais directly, pulling exact entries while using 2,900 tokens per query instead of 24,500 in standard search benchmarks.
Recall is intentionally binary. If a keyword misses or contains a typo, the engine yields an empty response rather than nearest neighbors. You must know your own tags, which prevents hallucinated or loosely matched context.
What it helps with
- retrieve stored file paths or credentials by exact keyword tags
- supply local context to coding agents with fewer tokens
- maintain personal knowledge in an archival plain-text index
- organize external files without copying or moving them
Example prompts for the agent
- Look up our staging tunnel configuration in ais under dev and proxy
- Store the path to this repository in ais with tags backend auth
- Query ais for notes tagged deployment checklist
How to install ais
Install
The MCP registry has no launch command for it: see the repository README
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about ais
How do you connect ais to Claude Code?
The agent hooks into the local CLI tool through a single configuration line. Built in C, the engine runs standalone without extra runtimes or external database servers.
What does ais do?
It records references to files, notes, and links under labels you choose. It returns exact matches when queried with those specific terms.
Why does ais avoid vector embeddings?
It skips semantic averaging to eliminate false positives. The query returns an exact match or nothing at all, preventing agents from receiving loosely related nearest neighbors.
