liliu-z/stashbase

Turn your local files into a Wiki for your agents. Open-source and local-first.

StashBase – a local‑first knowledge‑base for AI agents

What it is – StashBase is an open‑source desktop application (macOS, Windows, Linux) that turns the files on your computer into a searchable, linked Markdown wiki that AI agents can use as context. It keeps your original documents untouched; the generated wiki pages are ordinary Markdown files you own.

Key capabilities

  • Automatic wiki creation – ask an integrated agent (OpenQuill by default, or Claude Code / Codex) to “build a wiki” from a folder. The agent writes an index.md and supporting pages, linking back to the source files.
  • Semantic search – meaning‑based retrieval over text extracted from PDFs, DOCX, images (OCR) and audio/video transcriptions. Keyword search works out‑of‑the‑box; semantic search uses embeddings from a provider you configure (OpenAI, OpenRouter, etc.).
  • Chat‑driven workflow – a built‑in chat window lets you converse with the agent, request wiki updates, ask questions about the material, and have the agent edit pages on the fly.
  • Local‑first data model – all source files and wiki pages stay on your disk. Only extracted text and vector indexes are stored by the app; you can delete a folder from the library without losing any original files.
  • MCP (Model‑Client‑Protocol) compatibility – the app exposes a configuration that external MCP clients can use, so you can connect other LLM‑backed tools to the same library.

How you get it

  • macOS (Apple Silicon) – brew install --cask liliu-z/stashbase/stashbase
  • Windows (x64) – download the .exe from the latest release page.
  • Linux – download the .deb or AppImage for x86_64.
  • You can also build from source with Node 24+, pnpm, and Python 3.10+ (see the README for the full command list).

Typical workflow

  1. Open a folder containing notes, research PDFs, code docs, etc.
  2. In the chat, ask the agent to build a wiki for that folder.
  3. The agent creates wiki/index.md and additional topic pages, linking back to the originals.
  4. Search the library (keyword or semantic) or ask follow‑up questions; the agent can open source files side‑by‑side with the chat and improve the wiki as you learn.

Why it matters – Many AI‑assistant tools require you to copy‑paste or upload documents. StashBase lets you keep everything locally, gives you a persistent, editable knowledge base, and provides the same vector‑retrieval experience you get from cloud services without moving your data.

Status & community – Early‑alpha release; the project is actively seeking feedback on agent workflows, search quality, and cross‑platform stability. Support is available via Discord and the GitHub issue tracker.

License – Apache 2.0.

Related

  • Project
    shinpr/mcp-local-rag

    A local RAG system that indexes private documents (PDF, DOCX, Markdown, TXT) and provides hybrid semantic and keyword search via MCP or CLI without external APIs.

  • Project
    langflow-ai/openrag

    An intelligent agent-powered document search platform that enables users to upload, process, and query documents through a chat interface using RAG.

  • Project
    Fergana-Labs/stash

    A persistent knowledge base and shared memory layer for AI agents that connects to various data sources and records session transcripts to maintain context across runs.

  • Project
    BBC-Esq/VectorDB-Plugin

    A Windows-based tool for creating a vector database from documents, images, and audio files to enable Retrieval Augmented Generation (RAG) with various LLMs.

  • Project
    docling-project/docling-mcp

    An MCP server that provides AI agents with tools for converting PDFs into structured formats and generating new documents using the Docling library.