AI Gateway (LiteLLM Proxy)
The gateway routes requests to AI providers through an OpenAI-compatible API and manages access, spending limits, logging, and guardrails.
Find alternatives to AI Gateway (LiteLLM Proxy)
What it does
- Evaluation and observability
- Guardrails and safety
Features 5
- Virtual keys and budgetsPlatform teams create virtual keys and set budgets and rate limits, with cost shown by key, user, team and org.
- Python SDKThe litellm package calls 100+ LLM APIs in the OpenAI format with retries, fallbacks and cost tracking.
- MCP gatewayPuts all MCP tools behind one fixed endpoint with access control per key and team.
- Agent gatewayCalls A2A agents, Vertex AI Agent Engine, LangGraph and others with the same logs and access control as models.
- Admin UIDashboard where each team makes keys, adds models, sets budgets and reads its cost data.
Sources
- Pricing
-
"Free forever · self-hosted" for Open Source; Enterprise is "Talk to sales" and "Enterprise pricing is sized to your annual gateway request capacity, deployment architecture, and support needs".
- Description
-
Official documentation · 14 Oct 2026
Product page: "It gives one API to 100+ LLM providers"
- What it does
-
Official documentation · 14 Oct 2026
Product page: "It gives one API to 100+ LLM providers"
- Sold within
-
Sold on its own with published plans.
- Name
-
Official documentation · 14 Oct 2026
Name as published on the product page.
- URL
-
Official documentation · 14 Oct 2026
Product page, read 2026-10-14.
- Deployment
-
Official documentation · 14 Oct 2026
Product page: "Self-host anywhere, including fully air-gapped environments"
- Interfaces
-
Official documentation · 14 Oct 2026
Product page: "Self-host anywhere, including fully air-gapped environments"
Also from LiteLLM
-
LiteLLM Lens Application
LiteLLM Lens uses AI agents to analyze agent traces and group recurring failures.
-
LiteAdmin AI agent
LiteAdmin is a Slack agent for managing a LiteLLM gateway.
-
Auto Router Infrastructure service
Add-on for the LiteLLM AI Gateway that classifies each request and sends it to the model that should handle it, using heuristics, an LLM classifier or keyword rules, and reports the savings per request.
Something wrong here? Send a correction — quote this product id: litellm-ai-gateway-litellm-proxy.