Get Started

What is Trunkbase?

Trunkbase is a free AI companion app that turns the documents on your computer — PDFs, Word files, slides, spreadsheets, etc. — into high-fidelity context for your AI. While on-device search finds files, Trunkbase delivers context. It extracts and chunks file contents into formats native to LLMs, and serves your data via a built-in local MCP server so you can connect to your favorite AI apps — ensuring no files ever leave your machine.

Everything lives in your Vault — an ordinary folder you own — as Notes: clean, readable versions of your documents, stored as plain files. Changed a file? It updates. Damaged or deleted the index? It rebuilds. No database to manage, ever.

Untouched files in your local drive
Trunkbase
.MD .html .json .yml
AI-ready files in your local vault
Ollama
Claude
OpenAI
Gemini
Connect to your favorite AI

How does it work?

Creating contextual records of your data is called Ingestion. There are three strategies, mix them freely.

Manual Ingest

Record this file in now — Drag a file onto Trunkbase in Ingest mode, or use the file picker. It decomposes your file into a note, searchable within seconds. Drop a hundred files at once if you like — one bad file never sinks the batch, and your originals are never touched.

Watch a folder

Keep bringing files in, automatically — Point Trunkbase at a folder where documents naturally arrive — Downloads, a scanner's output, a project folder. Every new file flows into your vault automatically, even files that arrived while the app was closed. When a source file changes, its note updates in place. Originals in a watched folder are never moved, modified, or deleted.

Tunnel a folder

Index an external folder in place — Map an external folder into your vault as a paths-only index. Trunkbase records where each file lives and reads no content until you choose to — then decomposes entries on demand, one at a time or in bulk. Your original files stay exactly where they are, untouched.

Run built-in local MCP, Connect to AI

MCP (Model Context Protocol) is an open-source standard for connecting AI applications to external systems.Using MCP, AI applications like Claude or ChatGPT can connect to data sources (e.g. local files, databases), tools (e.g. search engines, calculators) and workflows (e.g. specialized prompts)—enabling them to access key information and perform tasks.

  1. Click the MCP chip in the status bar and turn on the Local MCP Server.
  2. Toggle the vault(s) you want to share — sharing is off by default, read-only unless you opt in, revocable in one click.
  3. Connect your app: Claude Desktop is a one-click connector bundle; Claude Code is a single command; any MCP-capable app works.

That's it. Ask your AI about your own documents, and it answers from your vault — grounded, cited, and entirely on your machine.