AnythingLLM Cloud Limitations

AnythingLLM Hosted Cloud is the quickest way to get a multi-user, managed, and hosted version of AnythingLLM on a custom domain.

While this form of accessing AnythingLLM there are some acute limitations you may not experience with other forms of AnythingLLM, like Desktop or self hosted.

No "built-in" LLM

AnythingLLM hosted cloud does not ship with a built-in LLM you can use like in our desktop instance. This is due to CPU limitations of the instance we provide for you, which has no GPU and limited CPUs and RAM.

Due to this, we limit access only to local LLMs that you can run yourself and connect to, or any supported cloud-based LLM provider.

Limited capacity for built-in embedder

️‼️

Beware! The built-in embedder will not block you from trying to embed a 5,000pg PDF, but it will crash your instance. (502 error).

AnythingLLM does allow you to use the built-in embedder model, which is extremely convenient, cannot embed on CPU any arbitrarily large document.

The Starter tier ships with very minimal compute resources while Professional ships with much more compute. This means that uploading a large document (in words, not file size) can overwhelm the CPU and cause the process to exit. This will result in a 502 error.

Issues with "Accuracy Optimized" Search in Workspace

Sometimes, your instance may become unresponsive or slow when using the "Accuracy Optimized" search in the Workspace if the workspace has a large number of files.

This is because the "Accuracy Optimized" search requires a lot of memory and CPU to run - which again are limited on the instance we provide for you.

You can prevent this by using the Default search mode in the Workspace settings.

No custom Agent supported

While AnythingLLM does support custom coded Agents, we do not support custom Agents in the hosted cloud due to security concerns as well other general limitations to running arbitrary code in a hosted environment.

If attempting to use a custom Agent, you will see an error about the system administrator not allowing custom Agents. This cannot be changed on hosted cloud.

If you need to use a custom Agent, you can use the AnythingLLM Desktop or a self-hosted AnythingLLM instance.

No MCP support

AnythingLLM does not support MCP (Model-Context Protocol) in the hosted cloud due to security concerns as well other general limitations to running arbitrary code in a hosted environment.

If you need to use MCPs, you can use the AnythingLLM Desktop or a self-hosted AnythingLLM instance.