Chat and agents
POST /v1/chat/completions with streaming, tool calling for agents and image inputs for vision.
← All productsEnterprise AI platform
AI Server
The all-in-one private AI server: chat, embeddings, images, speech and vision for every app and every desk, from hardware you control.
Run private OpenAI-compatible chat, embedding, image, speech, and vision APIs on your own infrastructure.
WindowsmacOSLinuxDockerKubernetes
Free to startWho it is for
Teams that need private chat, embedding, image, speech and vision APIs inside their chosen boundary.
What it helps you do
- Host models on Windows, macOS, Linux, Docker or Kubernetes
- Connect OpenAI-compatible applications
- Scale from one server to a gateway-managed farm
AI Server Team 5 — Store exclusive.One AI Server node and five AI Client seats for US$129.99 / month instead of US$199.94.
See team packs What it serves
One server for every AI job your team runs.
AI Server hosts the models once and serves them through familiar OpenAI-compatible endpoints. Every AI Suite app discovers it automatically, and any tool that accepts an OpenAI base URL can use it with an API key.
Embeddings for search and RAG
POST /v1/embeddings for document search, retrieval-augmented generation and classification.
Image generation
POST /v1/images/generations for text-to-image, image-to-image and inpainting.
Speech in and out
Text-to-speech, file transcription and live transcription over WebSocket.
Every AI Suite app
AI Client, AI Suite and the specialist apps find the server and offload their AI work to it.
Your own code and tools
Official OpenAI SDKs, LangChain, Semantic Kernel, Microsoft Agent Framework and LlamaIndex work unchanged.
For teams
One AI Server for the office. AI Client on every desk.
Put the models on one capable machine and give everyone else the lightweight AI Client: eleven AI workspaces for chat, writing, documents, translation, images, voice and dictation that use the server instead of each PC’s own hardware. Workstations stay light, models are downloaded once, and the data path stays inside your network.
Download models once
No multi-gigabyte model copy on every workstation; one server keeps the approved models warm.
Use the hardware you already have
Put the GPU in one machine and let laptops and thin desktops use it over the LAN.
Control who can call it
Network serving requires API keys that you issue and revoke on your own server.
AI Gateway
Grow from one box to a farm without reconfiguring clients.
AI Gateway is a mode of AI Server, not a separate product. It puts one endpoint in front of a pool of AI Server workers, so clients keep a single address while capacity grows.
Model-aware routing
Requests go to a worker that already has the requested model warm.
Failover
A failing worker is retried on a healthy one before the client notices.
Canary rollouts
Send a consistent slice of traffic to a new model or machine before trusting it fully.
Honest backpressure
When every worker is saturated, callers get a clear “retry shortly” instead of a timeout.
Zero-downtime upgrades
Drain a worker, upgrade it and return it to the pool while the endpoint keeps serving.
Deployment
Run it where your data already lives.
The same server and the same licence key work across every deployment shape.
Windows desktop
Install from Microsoft Store and serve your own machine in minutes; turn on network serving with Pro.
Docker
Run the softwaretailor/aiserver image with a mounted licence key, health probes and JSON logs.
Kubernetes
Deploy with Helm, scale GPU workers separately and route on readiness probes.
Azure and AWS
Bring-your-own-licence Kubernetes applications in Microsoft Marketplace and AWS Marketplace.
Editions
Free, Pro Personal or Pro Commercial?
Loopback is free; serving a network is Pro. The server enforces the edition itself, in every deployment shape.
| Capability | Free | Pro Personal | Pro Commercial |
|---|---|---|---|
| Serve AI Suite apps on the same computer | Unlimited | Unlimited | Unlimited |
| Other OpenAI-compatible clients (scripts, chat UIs, IDE plugins) | Rate-limited | Unlimited | Unlimited |
| Serve other devices on your network | — | Yes | Yes |
| Run as a background operating-system service | — | Yes | Yes |
| API keys for network clients | — | Up to 5 | Unlimited |
| Usage analytics | — | Yes | Yes |
| AI Gateway worker pool | — | — | Up to 25 workers, canary rollouts and automatic discovery |
| Rate limits, quotas and budgets, content moderation, audit signing, model lifecycle | — | — | Yes |
| AI Admin Console enrollment | — | — | Yes |
| Use rights | Local use on one computer | Private, household, hobby and educational use | Paid work and use by organisations |
| Price | Free | US$9.99 / month or US$99.99 / year | US$49.99 / month or US$499.99 / year per server node |
Enterprise adds unlimited estates, central entitlement allocation and priority support — talk to us. Prices are in USD before tax.
Privacy boundary
Run the service inside your chosen infrastructure.
The organisation places models and inference inside its configured environment. Network, endpoint, identity and model governance remain customer operating responsibilities.
Understand the full data pathDefault private pathOrganisation infrastructureCustomer-operated boundary
Privacy by architecture
Your prompts have nowhere else to go.
AI Server is not a proxy to a cloud API. Models run on your hardware, and requests travel from your apps to your server and back.
No vendor cloud in the data path
Prompts, documents and generated content stay between your clients and your server.
Content-free records
Usage analytics and audit logs record who called which endpoint and when, never the text of a prompt or response.
Your keys, your rules
API keys are issued and revoked on your server, not as accounts in someone else’s cloud.
Licensing without your content
Activation and periodic licence renewal send the licence key and device identifiers (installation ID, machine ID and a host-name hint), never prompts, files or outputs.
Start here
Try the real product before committing.
The useful free path has no time limit. Upgrade to Pro only when broader model choice or advanced capabilities earn their place.
Choose a Pro plan
Verified acquisition channelsSoftware Tailor online StoreMicrosoft StoreMicrosoft Marketplace (Azure)AWS Marketplace
Questions
AI Server questions, answered.
Is AI Server OpenAI-compatible? +
Yes. It serves /v1/chat/completions (with streaming, tool calling and vision), /v1/embeddings, /v1/images/generations, /v1/audio/speech, /v1/audio/transcriptions and /v1/models. Point any OpenAI SDK or tool at your server’s base URL and add an API key.
What does the free edition include? +
A complete local AI server for your own computer. Every AI Suite app on that computer uses it without limits; other OpenAI-compatible clients are rate-limited. Serving other devices on your network needs Pro.
Do I need AI Server to use AI Suite or the AI apps? +
No. AI Suite and the individual apps have a local AI engine built in. AI Server is for sharing models across several computers, serving AI Client, or calling private models from your own code.
How is AI Server licensed for a business? +
Choose Pro Commercial for paid work or any organisation, with one licence per running server; an AI Gateway in front of them needs no extra licence. One licence key activates every node you have paid for, on Windows, Docker, Kubernetes and the Azure and AWS marketplaces.
Does any of our data go to Software Tailor? +
No prompt, file or output passes through Software Tailor. Licence activation and renewal send the licence key and device identifiers (installation ID, machine ID and a host-name hint), and usage records are content-free.
Can we move from Ollama, LM Studio or another local server? +
Yes. Tools that accept an OpenAI-compatible base URL switch by changing that URL and adding an API key. Download the models you need again from AI Server’s Models page; the free edition lets you try the move on one computer first.
What hardware do we need? +
Start on the computer you have to validate the workflow; a CPU works for testing. Add a supported GPU for interactive chat, image and speech at team scale. The deployment planner sizes servers and seats for a workload.
What is the difference between AI Server and AI Gateway? +
AI Gateway is a mode of AI Server that puts one endpoint in front of several AI Server workers for routing, failover and upgrades. It is included with AI Server and is not a separate product.
Documentation
Set it up.
Deployment guides
See it in real deployments.
- One office, one AI box: serving every workstation from a shared server
- The zero-exposure default: one local AI server for everything on your machine
- Self-hosting AI in your own VPC: one container, env-configured, probe-ready
- Regulated enterprise AI: region policy, mandatory moderation, and signed audit
- The AI Server security model: what fails closed and where each key lives
- AI Gateway: one endpoint in front of a farm of private AI Servers
- Connecting an agent to AI Server: prove the whole tool-call loop
- Model-aware routing: different models on different boxes, one catalog to callers
- Zero-downtime upgrades: patching a private AI endpoint without a maintenance window
Continue exploring
View the catalogue Related products
Enterprise administrationFor IT
AI Admin Console
Identity, policy and evidence for an AI estate.
Windows 10/11 x64
Explore AI Admin Console
All-in-one private AIBest value
AI Suite
One desktop. Eleven private AI tools.
Windows
Explore AI Suite
Thin AI clientNew
AI Client
Eleven AI tools. One lightweight client.
Windows
Explore AI Client