- CollectionLLM Gateway & Routing
- Classificationagent-native
- MCPAvailable
- Agent SkillUnavailable
- InterfacesMCP · SDK · A2A · http
What it does
LiteLLM is an open-source AI gateway that unifies 100+ LLM providers behind an OpenAI-compatible API. Platform teams use it for virtual keys, spend tracking, budgets, rate limits, fallbacks, guardrails, and logging (Langfuse, OpenTelemetry, etc.). The Agent Gateway adds first-class A2A (Agent-to-Agent) routing, iteration budgets, trace grouping (X-LiteLLM-Trace-Id), and per-agent spend (X-LiteLLM-Agent-Id) for multi-step agents — see A2A docs.
Agent-facing interfaces
- Catalog protocols: MCP, SDK, A2A, http
- MCP: ✅ LiteLLM documents MCP integration on the AI Gateway (pass-through / proxy patterns for MCP servers — see docs index and search “MCP”). Treat as gateway feature rather than a single stdio package name.
- Agent Skill: ⚠️ No official AgentSkills repo located; search community:
- Onboarding type:
sdk
How agents onboard
pip install litellm
# Deploy proxy per https://docs.litellm.ai/docs/proxy/docker_quick_start
# Point OpenAI SDK base_url at your LiteLLM proxy (e.g. http://localhost:4000)
See the full onboarding section in the dossier.
Protocol surface
| Interface | Detail |
|---|---|
| HTTP | OpenAI-compatible chat/completions on proxy |
| A2A | POST /a2a/{agent_name}/message/send — see docs |
| Python | pip install litellm |
| Deploy | Docker, Kubernetes — see proxy deploy |