What's installed on your AnythingLLM server?
- AnythingLLM 1.19 with workspaces, document chat, AI agents and a developer API
- LanceDB, the built-in vector database
- The default embedder, which runs on the CPU and turns documents into vectors
- A single-password login, so the interface never opens without one
- Usage reporting turned off
- HTTPS through Caddy on an sslip.io address based on your IP
- Firewall with only ports 22, 80 and 443 open
What size do you need for AnythingLLM?
The official minimum is 2 GB of RAM, a 2-core CPU with AVX2 and 5 GB of disk, which is enough to store some documents and chat. The model runs elsewhere, so the real work on your server is processing documents. As a rule of thumb, look at 4 GB if you'll upload a lot of documents and 8 GB for heavy team use. Your documents and vectors decide the disk size. With no graphics card, large documents take longer to process on the CPU, and the official docs suggest an external embedding provider for that case.
| Use | Size | Price | Choose |
|---|---|---|---|
| Starter | 2 GB RAM · 2 vCore · 40 GB NVMe | $4.51/mo | |
| Medium | 4 GB RAM · 2 vCore · 60 GB NVMe | $6.89/mo | |
| Powerful | 8 GB RAM · 4 vCore · 120 GB NVMe | $13.62/mo |
Who is it for?
- Teams who want to search company documents, contracts or manuals by asking questions
- Websites that need a chat box answering from their own documents
- Teams who want a different model for each workspace
- Small organisations that need role-based access for staff
What happens after you order?
- Set the size and complete your payment.
- Your server starts and AnythingLLM installs itself.
- Open your server from My servers in the panel. On the Overview tab, wait until the App row says it's ready, then click Open app.
- Paste the value from the Password row into the login screen. There's no username, only one password.
- Choose a language model provider: enter your own API key, or connect to a model running on another server.
- Create a workspace, drag your documents in and start asking.
With every server
- Daily backups included in the price
- Unlimited traffic
- Full root access
- Scale RAM and CPU up or down later, and grow the disk
- Location: Istanbul · IST-1.
Good to know
- Your model provider's usage charges aren't included in the price. You manage the API key and the bill with the provider directly.
- If you choose a hosted provider, your questions and the relevant document passages go to it. To keep that data with you, connect a model you run yourself.
- In single-user mode, anyone with the password can read every chat and change every setting. Moving to multi-user mode can't be undone.
- If you change the embedding provider after processing documents, you'll have to delete and re-process them. Choose it with care at the start.
- To erase a document completely, delete it from My Documents. Removing it from a workspace leaves the parsed file and cache in place.
- Keeping AnythingLLM up to date is your job. The panel keeps showing the install password, so store any new one yourself.
AnythingLLM questions and answers
How much RAM does AnythingLLM need?
The official minimum is 2 GB of RAM, a 2-core CPU with AVX2 and 5 GB of disk. As a rule of thumb, go for 4 GB if you'll process many documents and 8 GB for team use. If that falls short, you scale up from your panel and keep the same data and IP address.
Does AnythingLLM work without a GPU?
Yes. In this install no model runs on the server, since the chat happens at the provider you connect. The AI work left on the server is the default embedder, which the official docs describe as CPU-only. On large documents that step can take a while.
Which models can I use with AnythingLLM?
You connect the AI provider you use with your own API key, or point it at an open model running on another server. You can also set a different model for each workspace.
Do my documents leave the AnythingLLM server?
Documents, vectors and chat history stay on your server. When you ask something, the question and the matching document passages go to the model provider you chose. If you connect a model you run yourself, that data stays on your own machines too.
Can I add more users to AnythingLLM?
Yes. Turn on multi-user mode in the settings, create the admin account, then add people with the Admin, Manager or Default role. Once you've made the switch you can't return to the single password, so decide up front.
Can I connect AnythingLLM to a separate vector database like Qdrant?
Yes. The default LanceDB runs inside the server, and the settings let you switch to another vector database such as Qdrant. Existing vectors don't move across: the official docs say you need to delete and re-embed each document.