LLM credentials & model compatibility

A2A Cloud routes user-provided model keys through LiteLLM. That gives chat and LLM-capable agents one OpenAI-compatible interface while preserving the user's chosen provider, model, usage, and pricing metadata.

How model setup works

In the dashboard, open LLM credentials and choose a provider. The model dropdown is populated from LiteLLM's public model cost map, so it includes the chat/completion models LiteLLM knows how to route and price.

Known providers, such as OpenAI, Anthropic, Gemini, Moonshot/Kimi, OpenRouter, Groq, Mistral, Cohere, xAI, DeepSeek, and Ollama, get a preset API base URL and default model. Additional providers discovered from LiteLLM still appear in the catalog, but you must enter the provider's OpenAI-compatible base URL before saving the key.

What agents receive

When an agent asks for ctx.llm, it receives credentials shaped like:

base_url=https://llm.a2acloud.io/v1
api_key=<scoped LiteLLM key or grant>
model=<platform model alias>

Agent code should call that endpoint as an OpenAI-compatible Chat Completions API. The control plane registers the user's real provider key behind the alias, so the agent does not need provider-specific routing logic.

Pricing and tracking

LiteLLM is the source of truth for usage and price attribution. The control plane registers provider-aware models, for example moonshot/kimi-k2.6 for Moonshot/Kimi or xai/grok-4 for xAI, so LiteLLM can match the model against its cost map.

If a model is in LiteLLM's cost map, token usage and spend are recorded with the provider's pricing. If a custom model is not in the cost map, the request can still run through the compatibility layer, but LiteLLM may not be able to attach provider pricing unless the gateway has pricing metadata for that model.

Examples

Moonshot/Kimi:

Provider: Kimi / Moonshot
API base URL: https://api.moonshot.ai/v1
Model: kimi-k2.6
Extra request body: {"thinking":{"type":"disabled"}}

OpenRouter:

Provider: OpenRouter
API base URL: https://openrouter.ai/api/v1
Model: openrouter/auto

Custom LiteLLM-compatible provider:

Provider: Custom / LiteLLM
API base URL: https://your-provider.example/v1
Model: provider/model-name

Use the dashboard dropdown when possible. It keeps the saved model aligned with LiteLLM's compatibility and pricing catalog.