Docker image
The best fit for most setups. Start it with one command; upgrades, migrations and rollbacks stay simple.
Run it on your own servers and connect cloud and local models through one gateway. Your data, keys and logs never leave your environment.
Latest version v2.3.0 (sample) · View changelog
All three ship the same features. The Personal Edition and commercial licenses use the same installer.
The best fit for most setups. Start it with one command; upgrades, migrations and rollbacks stay simple.
For servers that don't run containers. Installs as a systemd service that starts on boot.
For air-gapped networks with no internet access. Bundles the image and all dependencies.
* Installers haven't been released yet. Download links and SHA256 checksums will be published with the first public release. Version numbers are samples.
Grab a server that meets the system requirements and pick any install method.
8080 by default.http://localhost:8080Create the admin account on first visit to finish setup. On a remote server, use its IP address or domain instead.# The image name will be announced at release
docker pull <IMAGE_TBD>:2.3.0
# Start it and persist data to ./tokenhub-data
docker run -d --name tokenhub \
-p 8080:8080 \
-v $(pwd)/tokenhub-data:/data \
--restart unless-stopped \
<IMAGE_TBD>:2.3.0
# Open http://localhost:8080 to finish setup# docker-compose.yml
services:
tokenhub:
image: <IMAGE_TBD>:2.3.0
container_name: tokenhub
ports:
- "8080:8080"
volumes:
- ./tokenhub-data:/data
restart: unless-stopped
# Start from the same directory
docker compose up -d
# Open http://localhost:8080 to finish setup# The package download link will be announced at release
tar -xzf tokenhub-2.3.0-linux-<amd64|arm64>.tar.gz
cd tokenhub
# Installs a systemd service on port 8080
sudo ./install.sh
sudo systemctl status tokenhub
# Open http://localhost:8080 to finish setup* Sample commands. Placeholders such as <IMAGE_TBD> will be replaced at release; see the official install guide for final options.
Suggested specs (sample). Actual needs depend on concurrency, knowledge base size and log retention.
| Item | Minimum | Recommended (production) |
|---|---|---|
| Operating system | 64-bit Linux (x86_64 / ARM64) | Ubuntu 22.04, Debian 12, Rocky Linux 9 or another mainstream distro |
| CPU / RAM | 2 cores / 4 GB | 4 cores / 8 GB or more |
| Disk | 20 GB free | 100 GB SSD or more (grows with documents and logs) |
| Docker | Docker 20.10+ (container installs) | Docker 24+ with Docker Compose v2 |
| Browser | Latest Chrome, Edge, Firefox or Safari (for the admin console) | |
You only need GPUs on the servers that run local inference (such as Ollama or vLLM). If you only call cloud models, a regular server is enough. See the local model guide.
* Suggested values (sample). Final requirements will be listed in the release notes.
TokenHub is free, closed-source software. The Personal Edition and commercial licenses use the same installer; the only difference is how you use it.
For personal learning, research and non-commercial use
For companies, organizations and any commercial use
Specific rights and restrictions are governed by the Software License Agreement (coming soon).
Usually needs a commercial license
Usually free personal use
These examples are for reference only; the Software License Agreement has the final say. Not sure? Ask us.
No. TokenHub is free, closed-source software: you can download and use it at no cost, but the source code isn't published, and reverse engineering or redistribution isn't allowed. The Software License Agreement (coming soon) sets the exact terms.
Yes. Personal learning, research and non-revenue personal projects can use TokenHub free of charge.
Whenever TokenHub is used inside a company or organization, or as part of a commercial product or service. See "What counts as commercial use?" above; the license agreement has the final say.
TokenHub is bring-your-own-key (BYOK). You add your own API keys from cloud model providers, and you pay those providers directly; TokenHub never handles or marks up that spend. Local models such as Ollama or vLLM have no per-call fees; they just use your own compute.
No. TokenHub runs on your own servers, so prompts, knowledge base documents, logs and provider keys all stay in your environment. Only the requests you route to cloud models are sent to those providers.
Back up your data directory and read the changelog first. With Docker, pull the new image and recreate the container; your data stays in the mounted volume. With the Linux package, run the new version's install script. The official install guide will have the exact steps.
Download the offline installer, copy it to a server inside your network, load the image and run the install script. No internet access is needed. Pair it with local models (Ollama, vLLM) and a local embedding model so data never leaves your network.
Fill out the commercial licensing inquiry form or email hello@tokenhub.example. We'll learn about your use case, scale and deployment, then work out a license with you.
Free to download for personal use. Using it for business? Let's talk licensing.