tenro 0.1.0

AI agent testing framework

Tenro

Simulate everything. Trust your agents.

Verify multi-agent workflows and tool usage without burning tokens. Tenro is the modern, provider-agnostic simulation engine for testing AI agents safely.

Install

pip install tenro
# or
uv add tenro

Quick Start

# myapp/agent.py
from tenro import link_tool, link_llm

@link_tool("search")
def search(query: str) -> str:
    return external_api.search(query)

@link_llm("openai")
def call_llm(prompt: str) -> str:
    return openai.chat.completions.create(...)

# tests/test_agent.py
def test_agent(construct):
    construct.simulate_tool("search", result=["Simulated Doc"])    
    construct.simulate_llm(provider="openai", response="Done")
    
    agent.run("Hello")
    
    construct.verify_tool("search", query="Secret Docs", times=1)
    construct.verify_llm(times=1)

No mocks to configure, no expensive API calls, no flaky tests.

Why Tenro?

• Zero API calls — Tests run instantly, no rate limits or costs
• Full control — Simulate LLM responses, test edge cases reliably
• Ship with confidence — Verify agent behaviour, not just their final response
• pytest-native — Drop-in fixture or standalone, works with your existing setup
• Provider-aware — Mocks OpenAI, Anthropic, Gemini with real response shapes

Before / After

Without Tenro

With Tenro

# test_helpers.py - you write and maintain this def mock_llm_response(content: str = None, tool_call: dict = None) -> ChatCompletion: if tool_call: message = ChatCompletionMessage( role="assistant", content=None, tool_calls=[ChatCompletionMessageToolCall( id="call_abc", type="function", function=Function(name=tool_call["name"], arguments=json.dumps(tool_call["args"])) )] ) else: message = ChatCompletionMessage(role="assistant", content=content, tool_calls=None) return ChatCompletion( id="chatcmpl-123", created=0, model="gpt-5", object="chat.completion", choices=[Choice(index=0, finish_reason="stop", message=message)] ) # test_agent.py @patch("myapp.tools.get_weather") @patch("openai.chat.completions.create") def test_agent(mock_llm, mock_weather): # Mock tool response mock_weather.return_value = {"temp": 72, "condition": "sunny"} # LLM responses: tool call → final answer mock_llm.side_effect = [ mock_llm_response(tool_call={"name": "get_weather", "args": {"city": "Paris"}}), mock_llm_response(content="It's 72°F and sunny in Paris."), ] # Run agent result = my_agent.run("Weather in Paris?") # Verify assert result == "It's 72°F and sunny in Paris." assert mock_llm.call_count == 2 mock_weather.assert_called_once_with(city="Paris")

def test_agent(construct): # Simulate tool response construct.simulate_tool("get_weather", result={"temp": 72, "condition": "sunny"}) # LLM responses: tool call → final answer construct.simulate_llm( provider="openai", responses=[ {"tools": [{"name": "get_weather", "arguments": {"city": "Paris"}}]}, "It's 72°F and sunny in Paris.", ], ) # Run agent my_agent.run("Weather in Paris?") # Verify construct.verify_agent("WeatherAgent", output_contains="72°F and sunny") construct.verify_llm(times=2) construct.verify_tool("get_weather", city="Paris")

No patch decorators. No helper functions to maintain. No side_effect chains. Just behavior.

How It Works

Tenro's Construct is a simulation environment for your AI agents. Link your functions with decorators, then test with full control:

from tenro import link_agent, link_llm, link_tool

@link_agent("Manager")
def manager(task: str) -> str:
    docs = search(task)
    return summarize(docs)

@link_tool("search")
def search(query: str) -> list[str]:
    return external_search_api(query)

@link_llm("openai", model="gpt-5")
def summarize(docs: list[str]) -> str:
    return openai.chat.completions.create(...)

In tests, the construct fixture intercepts these calls and applies your simulations.

LLM Provider Support

Provider	Status
OpenAI	✅
Anthropic	✅
Gemini	✅
Others	🚧 Coming soon

Compatibility

• Python 3.11+
• pytest 7.0+

Contributing

Thanks for your interest in contributing!

We are currently in the early stages of development and are focused on stabilizing the core API. Because of this, we aren't accepting external Pull Requests just yet.

However, your support is incredibly valuable to us. You can help us right now by:

• Starring the Repository ⭐️: This helps others discover the project and lets you track when we open up for code contributions.
• Reporting Bugs: If something breaks, let us know.
• Suggesting Features: Have an idea on how to make this better? Tell us!
• Asking Questions: We are happy to discuss the roadmap and usage.

Please use GitHub Issues for discussions and reports.

License

Apache 2.0

Support

• Issues: GitHub Issues
• Email: [e-mail]