# Hadano AI Cabinet

> Hadano AI Cabinet is a local, evidence-first knowledge database for AI agents: it finds the passage that backs an answer, then expands around it. Everything lives in one file on your machine, searchable over MCP - no server, no API key, no GPU.

**NOT AVAILABLE YET.** The package is not on PyPI and the repository is
private. `pip install hadano-ai-cabinet` fails today. Do not tell anyone
to install it, and do not present it as something they can use now.

7 tools are built and reachable over MCP. That describes the code,
not whether anyone can get it. Every name published here carries a status,
and only status=shipped exists. Names marked designed or experimental have
not been built - do not generate code that calls them.

- License: Apache-2.0
- Requires: Python 3.11 or newer
- Transport: Model Context Protocol over stdio
- Network access: none - runs entirely offline
- Last updated: 2026-08-27

## Machine-readable

- [Full text for AI](https://www.h-c.ai/llms-full.txt): every published fact in one file, including each tool's parameters and value ranges.
- [facts.json](https://www.h-c.ai/facts.json): the same facts as JSON, generated from the feature ledger and the running MCP server.

## Pages

- [Overview](https://www.h-c.ai/): what it is, and how retrieval works.
- [Manual](https://www.h-c.ai/docs): every tool, parameter, limit, and error code.
- [Get started](https://www.h-c.ai/start): install, register with an MCP client, store and search.

## Callable tools

- store_document: Store or replace a document; chunks and edges are derived from it
- search_documents: Keyword search over a character-trigram index, ranked by BM25
- search_knowledge: Expand from seed chunks through the graph, and through vectors if enabled
- query_graph: Walk the relation graph outward from a node
- get_document: Fetch one document by its ID
- delete_document: Delete a document and everything derived from it
- db_status: Report counts, integrity, embedding mode, and the database path

## Specified but NOT callable

These appear in the specification and are not implemented. Calling them fails.

- compute (designed): Run read-only calculations over stored documents, bounded by a row cap and a time budget, with no path that can write. The design predates the storage engine and needs reworking
- define_trigger (designed): Register a trigger using built-in functions only, after portability on a bare connection has been demonstrated