AnythingLLM RAG Workbench
AnythingLLM 1.17.0 on Ubuntu 24.04 LTS, ready to run in your own AWS account. AnythingLLM lets you load documents and web pages into workspaces and ask questions about them with the LLM and embedding providers you choose (retrieval-augmented generation). Your documents stay on an EBS volume you control, and the UI listens on the instance itself: you open it through AWS Systems Manager port forwarding. This page covers what is on the image, the defaults it ships with and how billing works.
- AMI release
- 0.11.5 (AnythingLLM v1.17.0)
- Recommended
- t3.medium
- Storage
- 40 GiB+ encrypted gp3
- Access
- Session Manager, no inbound ports
- Released
- Oct 12, 2026
- Delivery
- Amazon Machine Image in your own AWS account
Chat with your own documents and web pages (RAG) using the LLM and embedding providers you choose, including Amazon Bedrock through its OpenAI-compatible endpoint.
Documents persist on your EBS volume, and each instance creates its own admin password on first boot.
The UI stays off the internet
It binds to localhost and you reach it through Session Manager port forwarding.
Quick start
Full guide →What is on the image¶
- AnythingLLM 1.17.0, as a Software Sushi container image built from the
upstream release and run by Docker as the container
sushi-anythingllm. - Chrome for Testing, pinned separately, for the built-in web page collector.
- A systemd service,
anythingllm, that starts the container at boot. - First-boot setup that creates this instance's admin password and the application's signing secrets. Nothing secret is baked into the image.
- Persistent storage for documents, settings and vectors in
/var/lib/anythingllm. - A walkthrough script,
/opt/sushi/docs/demo.sh, that checks the installation for you.
No provider keys or model weights are included. You configure your LLM and embedding providers in the UI after launch.
Defaults¶
| Setting | Default |
|---|---|
| UI | http://127.0.0.1:3001 (loopback only) |
| Sign-in | Single password, in /etc/anythingllm/admin.password, root only |
| Storage | /var/lib/anythingllm, kept across restarts and reboots |
| Telemetry | Off |
| Health | http://127.0.0.1:3001/api/ping |
Ports¶
| Port | Protocol | Listens on | Purpose |
|---|---|---|---|
| 3001 | HTTP | 127.0.0.1 |
AnythingLLM UI and API |
No inbound security group rule is needed. You reach the UI with Session Manager port forwarding.
Recommended instance¶
t3.medium (4 GiB of memory) with an encrypted gp3 root volume of at least
40 GiB and IMDSv2. Documents and vectors grow with use, so watch disk space.
The default vector database needs a CPU with AVX2; T3 instances provide it.
Billing¶
You pay an hourly software fee while the instance runs, as shown in the offer terms on the listing, which also shows any free trial. AWS bills EC2, EBS and data transfer separately. Your model providers bill their own usage.
Where to go next¶
- Getting started: subscribe, launch, sign in and connect a model.
- Configuration: files, settings and the password.
- Operations: security, service management, backups, updates and support.
- Troubleshooting: fixes for common problems.
Upstream project¶
AnythingLLM is developed by Mintplex Labs and published under the MIT license. Google Chrome terms apply separately to the bundled browser. See the AnythingLLM documentation for workspaces, providers, agents and the API. Software Sushi packages the image; Mintplex Labs does not endorse it, and Mintplex Labs and Google support are not included.
Last updated October 14, 2026 · Suggest an improvement