# RivetHub

> RivetHub is a self-hosted hub that records sessions, tool calls, and outputs from AI coding tools, and makes that history searchable by you and by the tools themselves. It is a distribution of the open-source RivetOS runtime.

If a human asked you to set up RivetHub, follow these steps. The prompt they may have pasted is:

Read https://rivethub.io/llms.txt and set up RivetHub on this machine.

## Set up on this machine

1. Confirm the host is Linux or macOS. Windows cannot run the hub installer. A Windows desktop client talks to a hub running elsewhere.
2. Fetch and run the laptop installer:

   `curl -fsSL https://get.rivethub.io/local.sh | bash`

   The installer can install Node.js 22+ if it is missing.
3. Verify the node with `rivetos local status`. The node answers at https://localhost:5174. On Linux the desktop app installs and finds localhost. On macOS the browser is the app for now. A desktop app is planned.
4. The installer wires capture and recall for supported tools it finds. If a tool is installed later, re-run the installer.
5. Confirm capture and recall from the installer output, then open the node.

Detail pages: [laptop quickstart](https://rivethub.io/install-local.html), [Omarchy](https://rivethub.io/omarchy.html), [server](https://rivethub.io/server.html).

## Availability and data boundaries

The local hub installer supports Linux and macOS. Windows has a client, not a native local-hub installer. There is an Android app. There is no iPhone app. Enrollment QR codes are for Android. The macOS desktop app is planned. Use the browser on macOS.

Shipped tools: Claude Code, Codex, Grok Build, Kimi Code, and Hermes.

Full-text and fuzzy search work without an embedding endpoint. Semantic search needs an optional embedding service. Collection is local, but configured agent, embedding, and compaction providers process the inputs sent to them. Do not infer that all processing is offline or that every integration is connected merely because it is supported.

Embedding and compaction endpoints and models can be configured on the server today. Choosing or changing those options from Desktop Settings → Memory is planned. Guided dataset curation and one-click SFT are planned and not shipped. Do not describe those desktop controls or export workflows as shipped features.

## Releases, support, and further reading

- [Current release manifest](https://rivethub.io/releases/latest.json): consult this for current versions, checksums, and app download filenames. Do not invent version numbers.
- [Release explanation](https://rivethub.io/releases.html)
- [Product and examples](https://rivethub.io/)
- [Applications](https://rivethub.io/apps.html)
- [Common questions](https://rivethub.io/#questions)
- [Server setup](https://rivethub.io/server.html)
- [Datahub setup](https://rivethub.io/install-datahub.html)
- [Add a node](https://rivethub.io/install-node.html)
- [Support](https://rivethub.io/support.html): GitHub issues, email rivetphilbot@gmail.com, or @rivetphilbot on X.
- [Privacy](https://rivethub.io/privacy.html): rivethub.io may record anonymous marketing-site pageviews. Self-hosted RivetHub / RivetOS do not phone home.
- [Runtime documentation](https://rivetos.dev/)
- [Runtime source](https://github.com/philbert440/rivetOS)