Nextcloud Assistant and Bee Flow: what each one does
· By Bee Flow · 4 min read
Nextcloud's Assistant and Bee Flow overlap in places and can run on the same Nextcloud. What each one is, where they differ, how they work together, and when the Assistant is all you need.

Administrators who look at Bee Flow often run Nextcloud's Assistant already, or have tried it. The two overlap in places, and they can run on the same Nextcloud at the same time. This is how we see the line between them. Nextcloud's side is taken from its own manuals for Nextcloud 35.
What Nextcloud Assistant is
The Assistant is the user interface for Nextcloud's AI features: tasks, the smart picker and Chat with AI. You choose the backend app behind each feature, and the manual lists them per feature: llm2 or the OpenAI and LocalAI integration for text and chat, stt_whisper2 for speech-to-text, translate2 for translation, and so on. Apps schedule their AI work through the Task Processing API, so a provider you register serves more than the Assistant window.
Two apps go further. Context Chat indexes users' files into a vector database so they can ask questions about their documents. It needs Nextcloud 32 or later, a deploy daemon, a backend ExApp whose major and minor version match the app, and hardware of its own (the manual recommends a dedicated machine).
Context Agent lets the Assistant act in Nextcloud, with tools for Calendar, Tasks, Contacts, Files, Talk, Mail, Deck, Forms, Share and Tables. It needs AppAPI 3.1.0 or later and a model that can call tools. It can also use remote MCP servers over streamable HTTP.
What Bee Flow is
Bee Flow is a separate server with its own database. It holds chat, agents, knowledge bases, automations and meeting notes, with a PII filter (Privacy Shield) in front of every model and a Compliance Center for the registers auditors ask for. The Bee Flow app for Nextcloud is an ExApp that shows Bee Flow inside Nextcloud. It works under each user's own Nextcloud account, and no password leaves Nextcloud. Models can be in the cloud, EU-hosted or on your own hardware. The server is fair-code, and the connector is AGPL.
Where they overlap, where they differ
| Nextcloud Assistant | Bee Flow | |
|---|---|---|
| What you install | Apps in Nextcloud, plus ExApps for local backends | A separate stack, plus the connector ExApp |
| Questions about documents | Context Chat, over Nextcloud files | Knowledge bases from files, web pages, tables and meeting notes, with citations |
| Acting in Nextcloud apps | Context Agent tools | Built-in tools for Files, Calendar, Contacts, Mail, Talk, Deck, Tasks, Notes, Tables, Forms and more |
| Other systems | Remote MCP servers in Context Agent | Google Workspace, Microsoft 365, GitHub, MCP servers and others |
Bee Flow also runs automations: a new file, a tag, a calendar change or a submitted form in Nextcloud can start one, as described in Nextcloud events that start automations. And one Bee Flow server can host several organisations.
Running both
The connector registers Bee Flow as a Task Processing provider for the text task types: summary, headline, topics, proofread, reformulate, simplify, formalise, improve, translate and free-form prompts. Under Administration settings → Artificial intelligence you pick Bee Flow per task type, next to the other providers you have. Those tasks then run on the model your Bee Flow organisation chose, with the Privacy Shield in front. Image generation, text-to-speech and transcription are not registered, so those stay with the providers you already use for them.
The other direction goes over MCP. Bee Flow serves an MCP endpoint at /mcp, and Context Agent's MCP Config takes remote servers. Add Bee Flow there with a token, and the Assistant can call Bee Flow's tools and the automations a user published as agent-callable. The token you enter there is used for every user, so a service set up this way acts as whoever created the token. Our Nextcloud install docs describe both steps, and An MCP server for Nextcloud compares the MCP options.
When the Assistant is all you need
If what you want is AI inside Nextcloud's own apps, on a model you host, the Assistant with llm2 or an OpenAI-compatible runtime is the shorter path. There is no second server to run and back up. Context Chat covers questions over Nextcloud files, and its manual describes which files it indexes.
Bee Flow is worth adding when you need more around it: automations on Nextcloud events, knowledge bases built from more than Nextcloud files, a PII filter before a cloud model, or Google and Microsoft accounts next to Nextcloud. It also asks more of you. The server needs at least 2 CPU cores and 4 GB of memory on top of Nextcloud, about 7 GB more with the PII guard and topic classifier, and a backup of its own, because your Nextcloud backup does not include it.
What an AI assistant can do inside Nextcloud shows the Bee Flow side in practice. For a local model on either side, see AI in Nextcloud with a local model, and self-host or hosted covers where Bee Flow runs.
To try it on your own Nextcloud, install Bee Flow next to it, or contact us if you would rather have us host it or want to talk it through first.