Ingate Documentation

Everything you need to integrate, configure, and operate the Ingate AI gateway.

New to Ingate?

Get your first LLM request routed through Ingate in under 5 minutes.

Go to Quickstart →

What Ingate Does

Ingate is an LLM gateway that sits between your applications, agents, and LLM providers. Change one URL, keep all your existing code. Ingate auto-detects the target provider from the request path. OpenAI, Anthropic, Azure, Ollama, and others work out of the box.

Unlike SDK-based observability tools that require code instrumentation, Ingate operates at the network layer as a transparent proxy. You get full observability, prompt management, an eval engine with datasets and a comparison playground, provider fallback, a skill marketplace, trace exports, and enterprise governance, without touching your application code.

Two ways to run Ingate

Use Ingate Cloud (sign up at app.ingateai.com, point at https://api.ingateai.com) or self-host the single Go binary or Docker image against PostgreSQL. See Deployment.

Architecture

Your Apps & Agents  →  Ingate  →  OpenAI / Anthropic / Azure / Ollama
                        ↓                    ↓
        Logs · Evals · Prompts · Skills    Exports → Phoenix / Langfuse / webhooks

Integration Example

The simplest integration is a single line change. Ingate auto-detects that requests to /v1/chat/completions go to OpenAI, so no provider header is needed:

tsapp.ts
// Before: direct to OpenAI
const client = new OpenAI({ apiKey: "sk-..." });

// After: route through Ingate (one line change)
const client = new OpenAI({
  baseURL: "https://api.ingateai.com/v1",
  apiKey: "sk-ingate-your-key",
});

// That's it. Your existing code works unchanged.
// Ingate auto-detects OpenAI from the /v1/... path.

Works with any provider

Anthropic, Azure, Ollama, and other providers are auto-detected the same way. For providers Ingate doesn't recognize, add an X-Ingate-Provider header or mark a provider as default. See Providers.

Agents

Ingate is built for agents as much as applications. Coding agents like Claude Code, Cursor, and Aider route their LLM traffic through the proxy for usage tracking, and manage the gateway itself over MCP. An agent with no credentials can connect to /mcp and call the ingate_setup tool to create an org, app, provider configs, and API key in one round trip.

bashterminal
# Track Claude Code usage: point it at Ingate
export ANTHROPIC_BASE_URL="https://api.ingateai.com"

# Let Claude Code manage the gateway over MCP
claude mcp add --transport http ingate https://api.ingateai.com/mcp \
  --header "X-Ingate-Key: sk-ingate-your-key"

See MCP Server for per-agent setup and Coding Agent Tracking for the proxy-based approach. Agents can also discover the API at runtime via /llms.txt and /openapi.yaml.

Next Steps