The first. The origin. The place a document opens and a thought begins.
A knowledge base that lives. Self-hosted. Open source. Off by default, opt-in for the rest. The server remembers, you think.
↓ scrollEvery file ships with its source. Read the script that runs the server. Read the change that adds a feature. Read the bug report that says "we got it wrong." AGPLv3-or-later, with a CLA so we can evolve the product cleanly.
No telemetry. No analytics shipped with the binary. No third-party scripts. The cloud features (AI suggestions, share links, Kaizen score sync) are off until you turn them on. The German data protection officers wrote us a thank-you note. We are not joking.
Orimora stores documents as a network. Every change is a pulse. Every pulse keeps the network breathing. Search goes through everything you've written — not just titles — and finds the note you wrote in 2023 that suddenly matters again.
A writer. A researcher. A founder. You keep notes for yourself because the alternative is forgetting. Orimora remembers so you can think.
Markdown. Local-first search. A private Kaizen score. One person, one server, one network that grows for years.
Three people. A shared doc. A comment thread that does not end in "but Slack?" Orimora gives your team one place where the conversation survives the meeting.
Permissions, audit trail, shared workspaces, comment threads that stay with the document, not with the chat window.
A document network your AI can read. The MCP server exposes your docs as resources. Claude reads them. You ask. It answers. You write. It edits.
The pulse keeps going. The opt-in is yours. The data stays on your machine, not in a model provider's context window.
The first. The origin. The place a document opens and a thought begins.
Delay. Wait. Stay. The slowness that lets a good sentence become a true one.
Open source. Open data. Open by default. Opt-in for everything that isn't.
Your existing database, or a fresh one. Orimora is a schema, not a service. You can `pg_dump` the whole thing and walk.
Expose your docs to Claude, Cursor, Windsurf, Continue. The agent reads, you stay in control. BYOK for every model provider.
One compose file. One port. Done. Or bare metal if you want to feel the air moving across the disks.
Import in 90 seconds. Your existing vault becomes a network. The links stay, the cross-references get richer.
| slug | name | what it does |
|---|---|---|
| @orimora/plugin-kaizenv1.4 | kaizen | Daily pulse, score history, private streak. Off until you opt in. |
| @orimora/plugin-graphv0.9 | graph | 3D network view. Zoom to a node, see its parents, its children, its siblings. |
| @orimora/plugin-mcpv1.0 | mcp | Expose docs to Claude, Cursor, Continue. Model Context Protocol, your network as resources. |
| @orimora/plugin-journalv0.7 | journal | Day One and Obsidian importer. Walks your local vault, keeps the wikilinks, drops it in. |
| @orimora/plugin-exportv1.2 | export | PDF, DOCX, ODT, EPUB. Markdown stays markdown. The export is a courtesy, not a chain. |
| @orimora/plugin-webhookv0.4 | webhook | On every pulse, fire a webhook. Pipe it to n8n, to your CI, to the kettle. |
The whole product. Every feature. Self-hosted on your hardware, your cloud, your basement, your Hetzner box in Falkenstein. No feature gates. No upsell. No asterisks.
For teams that need a contract, an SLA, a named contact, a security review. Same code. Same features. A different conversation.
Every feature ships in the free edition. No feature gates. The enterprise tier is a conversation, not a paywall.
Is this just another Obsidian?
No. Obsidian is local-first and single-user. Orimora is a self-hosted server you can put on the LAN or the open internet — your team, your MCP clients, your phone all reach the same library, with permissions and an audit trail.
What does AGPL mean in plain German?
You can read every line, change every line, and run it for your team. If you change it and offer it as a service to others, the changes come back upstream. We think that is the right deal for a knowledge base.
Can I self-host on a Raspberry Pi?
For a personal notebook, yes. For a team, you will want at least 2 GB of RAM and an SSD. The reference deployment is a single Docker container plus Postgres.
Does it work in German?
Yes — German is the first language we built for. UI, docs, search, even the AI prompts (when opted in) are translated by humans, not machines.
Where does my data live?
Wherever you put it. The reference deployment uses a Hetzner box in Falkenstein, Germany. The cloud-managed version stays in the EU, full stop. Nothing leaves the EU by default.
Will there be a managed cloud?
Not yet. We are focused on making the self-host path excellent first. A managed offering is on the long-term roadmap once the open beta settles.
Self-hosted team knowledge base. Made with ◆ in Germany. The server remembers; you think. We mean it.