Flint is an open source, local-first knowledge base that turns your computer into a private second brain. It combines markdown notes, linked-note navigation, a visual graph, and an infinite canvas, with optional AI assistance that runs through local models instead of a cloud service. Your vault stays on your device, and no cloud account is required to use it.
Watch the 30-second repo signal
Key Takeaways
Flint is a local-first knowledge base: notes stay on your device, with no cloud account required.
It offers markdown notes with wiki links and tags, a force-directed graph view, and an infinite canvas for visual thinking.
Its AI agent is optional and works with local models through Ollama, so your notes can serve as the AI's memory without leaving your machine.
It runs on macOS, Linux, and Windows, is licensed under MIT, and installs with a one-line script on each platform.
Flint is still under active development, so expect rough edges alongside fast-moving features.
What Flint Actually Is
Flint describes itself as a secure, local-first knowledge base, and that phrase does most of the explaining. Unlike note apps built around syncing to a vendor's servers, Flint stores your vault locally. You can point it at a folder workspace, write in markdown, and keep everything on hardware you control. As an open source project, its code is public, which is exactly what open source means in practice: anyone can inspect it, and the community can contribute fixes and features.
The project is young. Its repository was created in April 2026 and, as of November 3, 2026, it has 149 stars on GitHub. The version badge on its README reads 2.0.2, and its most recent activity dates to the same day, a sign of a project still being shaped in public. That youth cuts both ways: you get a fast-moving tool, but you should treat it as software under active development, not a finished product.
Notes, Graph, and Canvas: Three Ways to Think
Flint's core is a markdown editor with live preview, wiki links written as `[[Note Name]]`, `#tag` support, auto-save, daily notes, and a split editor and preview mode. Wiki links are the connective tissue: link one note to another, and Flint can follow those connections.
Those connections become visible in the graph view, an interactive force-directed graph where node size reflects how many connections a note has. You can drag, zoom, pan, search, and filter by depth, which makes it easy to spot which ideas are central to your thinking and which are orphaned.
The canvas is the third mode: an infinite board for visual thinking with text cards, note cards, image cards, and frame groups. You can drag to connect cards with colored lines, and links between note cards render automatically. Zoom, pan, undo, and redo are all supported, so the canvas works as a freeform planning surface rather than a rigid diagram tool.
The Local AI Angle
Flint's most distinctive feature is its optional local AI agent. It currently works with Ollama only, for example with `ollama pull llama3.2`, and it uses your notes as memory. It can read graph connections for extra context, search Wikipedia when internet access is enabled, and perform note-editing actions for supported requests.
This is the privacy story in concrete terms. With a cloud AI note app, your notes typically travel to a server to be processed. With Flint's approach, the model runs on your machine through Ollama, so the notes-as-memory loop never has to leave your device. Python 3 is optional and only needed for the local AI agent, while Node.js 18 or newer is needed only when building from source or running in development mode.
The README is honest about the state of this feature: several AI items are still marked as in progress, including using local models to answer questions from your notes, graph links, and optional internet search. Treat the AI agent as a promising work in progress, not a finished assistant.
Flint at a Glance
Aspect | Detail |
|---|---|
License | MIT |
Language | TypeScript |
Platforms | macOS, Linux, Windows |
Storage | Local vault, folder workspace support, no cloud account required |
Note format | Markdown with wiki links and tags |
Views | Editor, graph, infinite canvas |
AI | Optional local agent via Ollama |
Install | One-line script on Linux, macOS, and Windows PowerShell |
GitHub stars | 149, as of November 3, 2026 |
How to Try It
Installation is a one-liner. On Linux and macOS, the installer script checks for official prebuilt desktop binaries for a fast install, and falls back to configuring a desktop runtime when binaries are not available for your platform. On Windows, the same flow runs through PowerShell. An uninstaller script is provided too, runnable as `flint-uninstall` from the terminal.
A sensible first test: create a vault, write five notes linked with wiki links, open the graph to see the connections, then try the canvas with a few cards. If you want the AI features, install Ollama, pull a small model such as llama3.2, and see how the agent answers from your notes. That loop, notes plus graph plus local model, is the whole idea.
Who Flint Is For
Flint fits people who want a second brain without handing their notes to a cloud service: students building a study vault, developers keeping a technical knowledge base, writers linking research notes, or anyone curious about local AI. If you need polished mobile sync, team collaboration, or a mature plugin ecosystem, Flint is probably not there yet. If you value privacy, markdown files you truly own, and a graph of your own ideas, it is worth a look.