Skip to content

DeltaLLM

DeltaLLM gives your applications one place to call AI models. It connects to providers such as OpenAI, Anthropic, Gemini, and Bedrock, while your team controls access, routing, spending, safety, and monitoring.

Start with Docker See how DeltaLLM works

Choose what you want to do

I want to… Start here
Try DeltaLLM on my computer Get started
Connect an application to AI models Build with DeltaLLM
Manage models, keys, teams, and spending Administer DeltaLLM
Run DeltaLLM in production Deploy and operate
Understand the main ideas Understand the system
Look up an API, setting, or provider Reference
Watch short video walkthroughs Video tutorials

The shortest path to a working request

  1. Start DeltaLLM with Docker.
  2. Add a model.
  3. Send a test request.
  4. Create a key for your application.

You can usually complete these steps in a few minutes if you already have a provider API key.

What DeltaLLM manages

Your applications             DeltaLLM                    AI services
┌─────────────────┐      ┌─────────────────────┐      ┌─────────────────┐
│ OpenAI SDKs     │─────▶│ Access and limits   │─────▶│ OpenAI          │
│ HTTP clients    │◀─────│ Routing and safety  │◀─────│ Anthropic       │
│ Agent tools     │      │ Cost and monitoring │      │ Gemini and more │
└─────────────────┘      └──────────┬──────────┘      └─────────────────┘
                                    │
                         PostgreSQL and Redis

Applications call a public model name. DeltaLLM chooses the right provider deployment, applies your rules, sends the request, and records the result. Read What happens to a request when you need more detail.

Before production

The Docker quickstart is for learning and testing. A production deployment also needs protected network access, reliable PostgreSQL and Redis services, backups, monitoring, and a safe upgrade process. Use the production checklist before serving real traffic.

Get help