> ## Documentation Index
> Fetch the complete documentation index at: https://docs.niadra.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Microsoft Agent Framework

> The customer's memory as a Microsoft Agent Framework context provider, AutoGen's successor.

`NiadraContextProvider` is an Agent Framework context provider: before each run it adds the pack to the instructions and the history tools to the agent, and records the turns. Python only.

## Install

```sh theme={null}
pip install 'niadra[agent-framework]'   # agent-framework-core 1.19 or newer, below 2
```

## The five primitives

| Primitive    | How the adapter wires it                                                                                                                                                              |
| ------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Context      | Before each run the provider adds the pack (after the agent's own notes, with `agent_memory=`) and the `turn_block` as instructions, which the framework places after the agent's own |
| Turns        | The run's new user messages are the customer's turns; the run's response is the agent's turn, with its usage (input tokens, cache reads and writes) and model                         |
| Tools        | `FunctionTool`s whose input model is the kit's JSON Schema, bound to the customer                                                                                                     |
| Verification | `conversation.verify()` before the run                                                                                                                                                |
| Handoff      | `transferred_to_agent()` and `transferred_to_human()` record the transfer, for a workflow that hands over or an escalation to a person                                                |

## Minimal example

```python theme={null}
"""A Microsoft Agent Framework agent with the customer's memory as a context provider."""

import asyncio

from agent_framework import Agent
from agent_framework.openai import OpenAIChatClient

from niadra import AsyncNiadra, phone
from niadra.integrations.agent_framework import NiadraContextProvider

niadra = AsyncNiadra(channel="chat")


async def main() -> None:
    provider = NiadraContextProvider(niadra.conversation("thread-81", subject=phone("+5511912345678")))
    agent = Agent(
        client=OpenAIChatClient(), instructions="You are Acme's agent.", context_providers=[provider]
    )
    print((await agent.run("Where is my replacement lid?")).text)
    await niadra.close()


asyncio.run(main())
```

The same code is in `examples/agent_framework_agent.py`.

## Agent memory

`NiadraContextProvider(conversation, agent_memory=True)` puts the agent's own notes before the customer's context and adds the two tools. See [Agent memory](/en/concepts/agent-memory).

## Limits

* This is the adapter for the Agent Framework, AutoGen's successor in Microsoft's line. AutoGen's community continuation, AG2 1.x, has an [adapter of its own](/en/integrations/ag2); an AutoGen 0.4 agent uses the kit's tools and `conversation.context()` directly.
* Nothing here fails the run: without a pack, the instructions stay yours alone.
* Tested against `agent-framework-core` 1.19 with the model replaced by a fake and Niadra on the emulator.

## Next steps

<CardGroup cols={2}>
  <Card title="OpenAI and Azure OpenAI" href="/en/integrations/openai">
    the client `wrap()`, for those calling the model directly.
  </Card>

  <Card title="Internal agents" href="/en/guides/internal-agents">
    tasks, objects and actions that close open items.
  </Card>
</CardGroup>