Hello everyone,
I recently built three Nextcloud apps. One I made because I genuinely needed it β nothing existing quite fit, so I built my own. The other two I made as experiments: things I thought might make Nextcloud a little more useful to have around, so I tried building them and see. All three are now on the App Store, AGPL-licensed, with sources on GitHub, and I use all three every day on my own instance. I would genuinely value your assessment of them.
RegiBase β the one I needed. I used Passman for a long time and was grateful for it, but for the way I work it never quite felt comfortable, so I set out to build the account manager I personally wanted. It works like a light database: design your own fields (or start from a template), and browse records as a list, a spreadsheet-style table, or cards, choosing per field where each appears. It works together with Contacts, Tables, Files, Notes and Calendar when they are installed. Secret fields can be encrypted client-side (AES-GCM) under a master password, so the server only ever stores ciphertext; sharing a collection wraps the key per recipient. I put real care into getting data in and out β import from CSV/JSON (e.g. Google Password Manager) or from Contacts, and full password-protected backup & restore β and there is an occ CLI too. Available in 12 languages.
FormulaBase β the first experiment. Save the formulas you reuse (markups, unit conversions, loan payments, mixing ratios), define each variable once, then type numbers and read the result live. The math runs entirely in your browser with a small expression parser β no eval, no server round-trip. By the nature of the app it ships with a large number of imported templates (over 3,000 across 63 categories), and I have not been able to verify every single formula β so if you spot a bug, I would be grateful if you let me know.
Talk-Bot β the second experiment. I originally built it because I wanted to use Claude Code from within Talk; once that worked, I added Google Gemini and any OpenAI-compatible endpoint while I was at it. It turns a Talk conversation into an AI chat, running entirely inside Nextcloud as a Talk in-app bot: no external daemon, no webhook host, no container, in a few hundred KB of PHP. It supports Anthropic and Google natively, plus any OpenAI-compatible endpoint (OpenRouter, DeepSeek, Mistral, Groq, OpenAI, or a local server such as Ollama, vLLM or LM Studio), and can optionally drive a CLI tool so a flat-rate subscription answers instead of per-token API billing. By default every user gets a sandboxed assistant with no tools and no server access; a second, elevated tier exists but stays off unless an admin deliberately turns it on. I have tested the Claude Code and gemini-cli paths fairly thoroughly, but the OpenAI-compatible side is not as fully exercised yet β so reports there are especially welcome.
All three support Nextcloud 30β34 (just verified on 34).
- RegiBase β RegiBase - Apps - App Store - Nextcloud Β· GitHub - ktec-nc-apps/RegiBase: A personal database for Nextcloud β design your own fields, browse records as list / table / cards, encrypt secret fields client-side, share, and integrate with Contacts, Tables, Files, Notes & Calendar. Β· GitHub
- FormulaBase β FormulaBase - Apps - App Store - Nextcloud Β· GitHub - ktec-nc-apps/FormulaBase: Save your own formulas and calculate instantly in Nextcloud β plus 3,000+ built-in templates across 63 categories. All math runs in the browser, no eval, no server round-trip. Β· GitHub
- Talk-Bot β Talk-Bot - Apps - App Store - Nextcloud Β· GitHub - ktec-nc-apps/Talk-Bot: Chat with Claude, Gemini or any OpenAI-compatible model inside Nextcloud Talk β a lightweight in-app bot, in a few hundred KB of PHP: no external daemon, no webhook host, no container. Β· GitHub
Most of all, I would like to hear how they feel in actual use. Where your feedback points at something that genuinely needs improving, I will do my best to take it on board. The apps are already in the shape I set out to build, so I am not looking to redesign them β but if you find bugs, or have ideas that make them better, I will gladly fold those in. And if these apps turn out to be useful to you, that would make me happy.
One last, honest note: I am a Japanese speaker, and I have leaned on AI to write this post. I am a little unsure whether I can give you responses that fully meet your expectations β but if you will bear with me, I will do my best to answer, and I truly look forward to your frank opinions. Thank you for reading.