Must have
Installation requirements
- The CPU must support the AVX2 instruction set.
Application hosting
Hosting AnythingLLM
AnythingLLM is an all-in-one AI workspace for chatting with your documents, with built-in embedding, agents and a choice of local or hosted models. The Docker version runs on a small CPU server; a GPU only matters if it runs the model too.
Requirements last checked against the project's documentation.
A sensible first server for AnythingLLM is 2 vCPU, 2 GB RAM and 20 GB of storage. The cheapest listed plan that fits is OVHcloud VPS-1 (US) at $4.54/month.
- Min vCPU
- 2
- Min memory
- 2 GiB
- Min storage
- 5 GiB
Understand the setup
Find the software and services needed to run the application.
See requirementsBudget your resources
Plan memory, storage and CPU for the way you will use it.
Open resource plannerBuild your shortlist
Compare plans for the starting size, then review the provider and billing terms.
See plansRecommended
Planning guidance
- For the built-in LLM, a GPU is strongly recommended; small models run on the CPU but respond much more slowly.
Minimum installation resources
The documented minimum covers AnythingLLM with a hosted model. Running its built-in LLM adds a few GB of disk per model, and a GPU is strongly recommended for that.
Where to start
A sensible first server for AnythingLLM
- vCPU
- 2
- RAM
- 2 GB
- Disk
- 20 GB
AnythingLLM's documented requirement is 2 GB of RAM, a 2-core CPU with AVX2 and 5 GB of storage. We start with 20 GB of disk, our estimate, to leave room for uploaded documents and their embeddings.
Our shortlist
Cheapest plans that meet this size
-
$4.54/mo list price
-
$5.00/mo list price
-
$10.00/mo list price
-
$10.00/mo list price
-
$10.00/mo list price
-
$10.00/mo list price
-
$11.88/mo list price
-
$12.23/mo hourly × 730
One cheapest plan per provider, based on each provider's standard list price. Check the plan page for introductory and renewal pricing before you buy.
Deploy with
Supported paths
The host must provide
Runtime
- Docker
- For the self-hosted server version
- Model provider
- A hosted API or a local model; the built-in LLM needs extra disk and ideally a GPU
Storage
Persistent data
Mount a host folder at /app/server/storage; without it, all data is lost when the container restarts. Documents, embeddings and any built-in models live there, a few GB per model.
Step by step
Plan an AnythingLLM deployment
-
Check the CPU and pick a model source
Confirm the server CPU supports AVX2, then decide whether a hosted API or the built-in LLM will answer questions. The built-in LLM needs extra disk and is much faster with a GPU.
Official instructions -
Run the container with persistent storage
Mount a host folder at /app/server/storage so documents and settings survive restarts. The web interface listens on port 3001.
Official instructions -
Lock it down before sharing
A new instance has no login. Turn on password protection for personal use or multi-user mode for teams, and use a server or a private network such as Tailscale instead of tunnelling a personal machine to the internet.
Official instructions
Security
Before you go public
- A fresh instance has no authentication; enable password protection or multi-user mode before anyone else can reach it.
- Do not tunnel a personal computer to the public internet to share AnythingLLM.
- Back up the storage folder, which holds documents, embeddings and settings.
Cost
Beyond the server price
- With a hosted model the server is small; token charges at the provider are the running cost.
- Running models locally moves the cost to GPU hosting and disk for model files.
Want the full monthly bill? Use the hosting cost calculator. Add storage, backups and other costs to your estimate.