Central LLM Wiki vs Hjarni
The closest neighbour: both are Markdown knowledge bases with MCP. The differences are whether anything compiles for you, and how you host it.
Choose Hjarni if
You want a clean hosted note store that many AI clients read and write, decide yourself (or via your agent) how each note is written, and need email-to-note and native mobile apps.
Choose Central LLM Wiki if
You want sources compiled into a wiki by rules, with the index and log kept current, plus bibliography, graph and lint, while controlling the deployment yourself.
| Feature | Hjarni | Central LLM Wiki |
|---|---|---|
| Mechanism | Storage: “nothing is automatic”, you or your AI write notes deliberately | Compilation: raw/wiki/schema layers with Ingest, Query and Lint built in |
| Agent access | MCP with OAuth or token; listed in the Claude and ChatGPT connector directories | MCP with token, or read-only OAuth 2.1 + DCR; paste the URL into Claude.ai/ChatGPT, no directory listing |
| Bibliography | Not offered | DOI/Crossref, BibTeX/CSL, [@citekey], researcher template |
| Graph and versions | No graph; 30-day trash versions | Graph (filters, timeline), a version per save, rollback |
| Mobile | iOS and Android apps | PWA (add to home screen); no native apps |
| Interface language | German not found | German and English |
| Deployment | Hjarni hosted service | Self-host on infrastructure you control with the full feature set |
| Export and source | Markdown bundle export; not open source | Markdown zip; Apache-2.0, self-hostable |
Trade-offs we admit
- Hjarni launched earlier; directory listings, email-to-note and native apps are things we do not have yet.
- Central LLM Wiki auto-ingest uses a model provider that you configure in your self-hosted instance; Hjarni does not run a model.
- The other product’s features and deployment details are as published on its site; “not found” means not found on the check date, not necessarily absent.
Get started on your own serverTheir facts checked on 2026-09-04 · Hjarni Website