When AnythingLLM Cloud is the better choice
When you want Mintplex Labs, the team that builds AnythingLLM, to run your instance and answer support tickets. Pro's 72-hour support SLA is something a server of your own does not come with.
AnythingLLM 1.16.1 AI
Free credit available Limited time
AnythingLLM hosting One workspace your whole team signs in to.
AnythingLLM hosting on Selfhost.dev puts the server version of AnythingLLM, with its built-in vector database, on a dedicated server we provision, secure and run. It opens at its own HTTPS address behind your password. When the team joins, switch on separate logins in settings and bring one model API key.
One container holds the app, its vector database and your documents. Telemetry is off from the first boot.
AnythingLLMone-click template ready
- AnythingLLM1.16.1, pinned
- vector databaseLanceDB, built in
- accessyour password, before the first boot
- telemetryoff
- modelsbring your own key
European 4 GB server$0.02/hr
In one line
What is AnythingLLM hosting?
What it is
AnythingLLM is an open source workspace for chatting with your own documents through any LLM, with agents and a vector database built in. AnythingLLM hosting runs its server version where your whole team can reach it.
Where to start. Open the console with AnythingLLM selected
Who it is for
Teams that want a private assistant over their own files, with one admin and a login for each person.
When to use it
When the Desktop app on one laptop is no longer enough. If you would rather pay AnythingLLM's own team to run it, their Cloud plans fit better.
Desktop vs server
AnythingLLM Desktop or a hosted server?
AnythingLLM Desktop is a single-user app: it lives on one laptop and stops when that laptop sleeps. The server version is the one AnythingLLM's own docs recommend when several people share it, reach it from a browser or need access rules per workspace. Here it runs on a server that stays on.
The stack, end to end
Callers
Your team in the browser
Edge
TLS and routing
Your project server
Container
AnythingLLM 1.16.1
Built in
LanceDB and SQLite
Platform secret
Your password
Outside the server
Your model provider
| AnythingLLM Desktop | AnythingLLM on Selfhost.dev |
|---|---|
| One user, on the machine it is installed on | Your team, from any browser |
| Reachable while that laptop is awake | Reachable whenever the server runs |
| No logins and no roles | One password, then a login and a role per teammate |
| Documents on that laptop's disk | Documents in one storage volume on the server |
The model runs at your provider; documents and vectors stay on this server. Deploy the stack
Containers
AnythingLLM Docker, set up for you
The AnythingLLM Docker quick start runs the latest image on port 3001 with an empty settings file. That instance starts with no password, so anyone who reaches it can add model keys and read every document. This template pins mintplexlabs/anythingllm 1.16.1 and sets your password before AnythingLLM first starts.
| Running AnythingLLM Docker by hand | AnythingLLM on Selfhost.dev |
|---|---|
mintplexlabs/anythingllm:latest, whatever that is today | mintplexlabs/anythingllm:1.16.1, an exact tag |
| No password until you set one in the app | AUTH_TOKEN from your password before the first boot |
| Plain HTTP on port 3001 | HTTPS with a certificate per hostname |
| Usage telemetry sent by default | DISABLE_TELEMETRY=true |
| Signing secrets made on first start | SIG_KEY, SIG_SALT and JWT_SECRET kept, so sign-ins survive a redeploy |
| A storage folder you create and mount | One storage volume for documents, vectors and the database |
Server
AnythingLLM hosting requirements
AnythingLLM hosting requirements stay modest because the model runs at your provider. AnythingLLM's own docs ask for 2 GB of RAM and 2 cores; this template starts on 4 GB, the smallest server we offer, which leaves room for embedding documents and for a second light app beside it.
What it needs
- AnythingLLM
- 1.16.1, pinned
- Vector database
- LanceDB, built into the same container
- Memory and CPU
- 4 GB, 2 cores; x86 or ARM
- Disk
- 40 GB on the 4 GB server, for documents and vectors
- Models
- None included; bring a provider key
- What a 4 GB server handles
- AnythingLLM plus up to four more light templates
Server size: 4 GB, enough for AnythingLLM. Deploy on a 4 GB server
Models
Bring a model: AnythingLLM ships without one
AnythingLLM answers with whichever LLM you connect, so this template includes none. Add one provider key in settings and every workspace can use it. Documents are embedded on the server by AnythingLLM's built-in embedder, so the key you add is only needed for chat and agents.
On the server (we set it up)
- AnythingLLM 1.16.1 and its LanceDB vector database
- The built-in embedder for your documents
- HTTPS, your password and the storage volume
What you bring
- One API key from OpenAI, Anthropic, Gemini or another provider AnythingLLM lists
- Or a gateway: add our LiteLLM template on the same server and choose LiteLLM as the provider
- Your documents
Quickstart
How to deploy AnythingLLM in 3 steps
One container, one password, no settings file to write.
-
Choose AnythingLLM and a 4 GB server
Launch opens the AnythingLLM template in the console. Pick a location and set a password of at least eight characters; it locks the instance from the start.
-
Launch it
The template deploys in about six minutes, most of it pulling the image. Then open your address and unlock AnythingLLM with that password.
-
Add a model, then your team
In settings, add your LLM provider key and upload a few documents. To give each person a login, switch on multi-user mode under Security.
Deploy AnythingLLMStep 1 starts in the console.
Launch AnythingLLM Billing
What AnythingLLM hosting costs
AnythingLLM hosting here is one server line. AnythingLLM is free under the MIT licence, the template adds no fee and teammates cost nothing extra. A European 4 GB server is $0.02/hr, about $15 to $16 a month, paid by the hour from prepaid credits. Your model provider bills you for chat separately.
| Item | Cost |
|---|---|
| AnythingLLM and its vector database | $0 |
| Teammates, workspaces and documents | $0, limited only by the server size |
| Bandwidth | $0, traffic is included in the server price |
| Optional daily backups | +20% of the server price |
| Chat and agent calls | Billed by your model provider to your own account |
| Server | RAM, cores, disk | Where | Rate | About a month |
|---|---|---|---|---|
cx23 (recommended) | 4 GB, 2 cores, 40 GB | Europe | $0.02/hr | $15.18 |
cax11 (ARM) | 4 GB, 2 cores, 40 GB | Europe | $0.02/hr | $16.35 |
cpx21 | 4 GB, 3 cores, 80 GB | US | $0.12/hr | $87.75 |
cpx22 | 4 GB, 2 cores, 80 GB | Singapore | $0.10/hr | $72.56 |
cx33 (more documents) | 8 GB, 4 cores, 80 GB | Europe | $0.03/hr | $23.36 |
Monthly figures assume 730 hours.
If your balance reaches zero
With a card on file or a past top-up, AnythingLLM keeps running for 5 days after your balance reaches zero. Then it is stopped and kept for 24 hours before it is deleted; a top-up in that time restarts it. After the delete, a copy of the server is kept for 14 days and you can restore the project from the Projects page. We email you when the balance reaches zero and again when it stops. Welcome-credit-only servers are deleted at zero, with a 3-day copy if the project went live.
Price another size or region in the calculator, no signup.
Options
AnythingLLM Cloud vs self hosted on Selfhost.dev
AnythingLLM Cloud is the vendor's own hosted plan: a private instance run by Mintplex Labs, starting at $50 a month; you still bring your own LLM API key. Self hosted on Selfhost.dev, the same open source app runs on a dedicated server you control, for the server's hourly price.
| Option | What you pay | What you get |
|---|---|---|
| AnythingLLM Desktop | Nothing | One user on one machine |
| AnythingLLM Cloud Basic | $50 a month | A private instance and a custom subdomain; bring an LLM key |
| AnythingLLM Cloud Pro | $99 a month | Higher performance and a 72-hour support SLA |
| AnythingLLM on Selfhost.dev | $0.02/hr for 4 GB in Europe | A dedicated server, your own domain if you add one; bring an LLM key |
Cloud figures from anythingllm.com/cloud.
Status
Five checks on every AnythingLLM server
- Never open on the internet. The password is set before AnythingLLM first starts, so the address is locked from its first request.
- A stalled deploy is retried automatically, not abandoned.
- A deploy is never reported done until the URL answers.
- Your AnythingLLM version stays put. The image is an exact tag; a restart or redeploy brings back the same build. When we ship a newer release, the console offers it as an upgrade.
- Sign-ins survive a redeploy. The signing secrets are kept, so nobody on your team is logged out by one.
~6 min to deploy
The console estimates about six minutes for the AnythingLLM template after the server is ready; first-time server provisioning adds time.
Notes
AnythingLLM hosting limitations
- No model included. Add a provider key or a LiteLLM gateway in settings; nothing is preloaded or billed through us.
- The disk grows with the server. Documents and vectors share the server's disk; Scale up now in Project Settings moves to a bigger size with its data, after a short downtime.
- One server, one copy. No second instance to fail over to; nothing is replicated beyond the optional daily backups.
Nothing here in your way? The template deploys in about six minutes.
Launch AnythingLLMBackups
AnythingLLM backup and restore
AnythingLLM backup and restore is simple because everything lives in one storage volume: uploaded documents, the LanceDB vectors and the SQLite database with users and chat history. The server can take an optional daily image of all of it; to copy the volume out, ask us for container access over SSH.
| Method | What it covers | Where it lives |
|---|---|---|
| Daily server backup (optional, +20%) | AnythingLLM and its whole storage volume; the last 7 kept | With the server; deleted with it |
A copy of anythingllm-storage | Documents, vectors, users and chats | Your storage, via container access over SSH, granted on request |
Sizing
Not sure 4 GB holds your documents?
Tell us how many people will use it and roughly how many files you plan to upload. We send back the size we would pick and what it costs a month.
Help
Frequently asked questions
Is AnythingLLM free?
Yes. AnythingLLM is open source under the MIT licence. Both the Desktop app and the Docker version cost nothing. What costs money is where it runs and the model it calls. Here you pay for the server by the hour; your model provider bills you for chat. See pricing.
Can several people use one AnythingLLM?
Yes, on the server version. It starts in single-user mode behind the password you set in the console. Turn on multi-user mode in settings and that password gives way to real accounts: you become the admin and each teammate gets their own login, with access set per workspace.
Does AnythingLLM need a GPU?
Not here. The language model runs at the provider you connect, so the server only stores documents, embeds them and serves the app. Our servers have no GPU, which is why this template ships without a local model. A 4 GB server with two cores covers the app itself.
Can I use my own domain for AnythingLLM?
Yes. Every deployment gets its own HTTPS address first. To use a domain you own, point its DNS at the server, then add it from the project in the console; a certificate is issued for it and AnythingLLM answers there. Your team signs in at that address.
Your documents, your team, your server. Locked from the first request.
- AnythingLLM with LanceDB built in and telemetry off
- Your password before the first boot, logins per teammate after
- A European 4 GB server, about $15 to $16 a month
About Selfhost.dev
Selfhost.dev is a managed self hosting platform for developers and small teams. Open source apps and databases on dedicated servers you control, billed by the hour, in six locations across three continents.
Best for: AnythingLLM, LiteLLM, Open WebUI and managed PostgreSQL, MySQL, Redis, ClickHouse, OpenSearch and Kafka.