TinyAgent

As part of the bare bones library, we provide our own Python implementation based on HuggingFace Tiny Agents.

You can find it in any_agent.frameworks.tinyagent.

Examples

Use MCP Tools

from any_agent import AnyAgent, AgentConfig
from any_agent.config import MCPStdio

agent = AnyAgent.create(
    "tinyagent",
    AgentConfig(
        model_id="mistral/mistral-small-latest",
        instructions="You must use the available tools to find an answer",
        tools=[
            MCPStdio(
                command="uvx",
                args=["duckduckgo-mcp-server"]
            )
        ]
    )
)

result = agent.run(
    "Which Agent Framework is the best??"
)
print(result.final_output)

Experimental any-llm Support

TinyAgent has experimental support for the any-llm library, which provides a unified interface to different LLM providers. To enable this feature, set the USE_ANY_LLM environment variable:

export USE_ANY_LLM=1

You'll also need to install the any-llm dependency:

pip install 'any-agent[any_llm]'

When enabled, TinyAgent will use any-llm instead of LiteLLM for model completions

from any_agent import AnyAgent, AgentConfig
from any_agent.config import MCPStdio

# Set environment variable to enable any-llm
import os
os.environ["USE_ANY_LLM"] = "1"

agent = AnyAgent.create(
    "tinyagent",
    AgentConfig(
        model_id="ollama/mistral-small3.2:latest",
        tools=[
            MCPStdio(
                command="uvx",
                args=["duckduckgo-mcp-server"]
            )
        ]
    )
)

result = agent.run(
    "Find me one new piece of AI news"
)
print(result.final_output)

See the any-llm documentation for a complete list of supported providers.