Platform extensions let agents access external services and UI components via dependency injection. They provide configured services automatically, make agents portable, support user-configured services, handle authentication, and remain optional.

​
Extension Types

​
Service Extensions

Provide access to external services with automatic configuration:
  • LLM Service: Access to language models (OpenAI, Ollama, IBM Granite, etc.)
  • Embedding Service: Text embedding generation for RAG and similarity search
  • Platform API: File storage, vector databases, and platform services
  • MCP: Integration with Model Context Protocol tools

​
UI Extensions

Enhance the user interface with rich components:
  • Trajectory: Step-by-step agent reasoning visualization
  • Citation: Source references with clickable links
  • Form: Interactive forms for structured user input

​
Dependency Injection Pattern

Extensions are injected using Annotated type hints: 
from typing import Annotated
from beeai_sdk.a2a.extensions import LLMServiceExtensionServer, LLMServiceExtensionSpec

async def my_agent(
    input: Message,
    context: RunContext,
    llm: Annotated[LLMServiceExtensionServer, LLMServiceExtensionSpec.single_demand()]
):
    # Platform provides configured LLM access
    pass

​
Extension Availability

Extensions are optional by design. Always check if provided: 
if llm:
    # Use LLM functionality
    pass
else:
    # Provide fallback or inform user
    yield "LLM not configured"

​
Model Suggestions

Service extensions accept suggested models in preference order: 
LLMServiceExtensionSpec.single_demand(
    suggested=("openai/gpt-4o", "ollama/llama3.1", "ibm/granite-3-8b")
)
Platform selects the first available model from suggestions. Extensions make agents adaptable to different deployment environments while providing rich user experiences.