Autonomous Agent — Python Sample

10/20/2026

This sample demonstrates a purely autonomous agent built with the Microsoft Agent 365 SDK for Python. It has no chat functionality — it runs entirely as a background service.

Every 60 seconds, the agent:

Prompts Azure OpenAI to fetch trending GitHub repositories using a registered tool
The model calls the get_trending_repositories tool, which queries the GitHub Search API
The model summarizes the results into a readable digest
The digest is logged to the console

All operations are manually instrumented with Agent 365 observability using the A365 Observability SDK tracing scopes (InvokeAgentScope, InferenceScope, ExecuteToolScope). This makes it a useful reference for instrumenting any non-interactive or custom agent loop.

For comprehensive documentation, visit the Microsoft Agent 365 Developer Documentation.

What This Sample Demonstrates

Pattern	Where
Autonomous background service with LLM tool calling	`github_trending_service.py`
GitHub Search API tool with observability	`tools/github_trending_tool.py`
Azure OpenAI chat completions with function calling	`github_trending_service.py`
A365 observability with manual tracing scopes	`github_trending_service.py`, `tools/github_trending_tool.py`
Microsoft OpenTelemetry distro with S2S exporter	`main.py`
3-hop FMI/FIC token flow for observability export	`observability_token_service.py`

Prerequisites

Python 3.11 or higher
Azure CLI
Agent 365 CLI (install: dotnet tool install --global Microsoft.Agents.A365.DevTools.Cli --prerelease)
An Azure OpenAI resource with a deployed model (e.g., gpt-4o)
An Entra tenant with at minimum the Agent ID Developer role

Environment Configuration

Agent 365 Setup

Log in to Azure: az login
Provision the agent:

cd python/autonomous/github-trending
a365 setup all --agent-name <your-agent-name>

Copy the provisioned blueprint and actual agent instance values from the setup output into your .env file (see below).

Verify eligible agent instance registration and authorization for permissionless S2S export. Creating an Entra identity or selecting the S2S endpoint alone is insufficient; Agent365.Observability.OtelWrite is not a universal prerequisite. Workload permissions remain independent. On 401/403, check IDs, blueprint credentials, instance registration and authorization rather than blindly adding OBS grants.

Configuration

Copy .env.template to .env and fill in the values from a365 setup all output:

cp .env.template .env

Variable	Set by	Description
`AZURE_OPENAI_ENDPOINT`	Manual	Azure OpenAI resource endpoint
`AZURE_OPENAI_API_KEY`	Manual	Azure OpenAI API key
`AZURE_OPENAI_DEPLOYMENT`	Manual	Model deployment name (default: `gpt-4o`)
`AGENT365_TENANT_ID`	CLI	Entra tenant ID
`AGENT365_AGENT_ID`	CLI	Agent identity ID (separate from blueprint)
`AGENT365_BLUEPRINT_ID`	CLI	Blueprint app registration ID
`AGENT365_CLIENT_ID`	CLI	Blueprint app ID (same as blueprint ID)
`AGENT365_CLIENT_SECRET`	CLI	Blueprint client secret
`AGENT365_AGENT_NAME`	CLI	Display name shown in traces
`AGENT365_AGENT_DESCRIPTION`	CLI	Agent description shown in traces
`AGENT365_USE_MANAGED_IDENTITY`	Manual	`true` for production (MSI), `false` for local dev
`ENABLE_A365_OBSERVABILITY_EXPORTER`	Manual	`true` to enable the A365 span exporter. Required for exporting traces to the A365 observability endpoint.
`ENABLE_A365_OBSERVABILITY`	Manual	`true` to enable SDK scope spans (`InvokeAgentScope`, `InferenceScope`, `ExecuteToolScope`). Required for the Python SDK.

Variable	Description	Default
`GITHUB_TRENDING_LANGUAGE`	Programming language filter	`python`
`GITHUB_TRENDING_MIN_STARS`	Minimum star count	`5`
`GITHUB_TRENDING_MAX_RESULTS`	Number of repositories per digest	`10`
`HEARTBEAT_INTERVAL_MS`	Polling interval in milliseconds	`60000`

The GitHub Search API is unauthenticated — no API key required (rate limit: 10 requests/minute).

Running the Agent Locally

Quick start (Azure OpenAI only)

You can run the agent with just Azure OpenAI credentials — no Agent 365 setup required. Create a minimal .env:

AZURE_OPENAI_ENDPOINT=https://your-resource.openai.azure.com/
AZURE_OPENAI_API_KEY=your-api-key
AZURE_OPENAI_DEPLOYMENT=gpt-4o

Then:

cd python/autonomous/github-trending
pip install -e .   # or: uv pip install -e .
python main.py

The agent will skip the observability token service and log a warning:

Agent365 credentials not configured — skipping token service.
Run 'a365 setup all' to enable A365 observability export.

To enable full A365 observability, complete the Agent 365 Setup steps and fill in the remaining .env values.

Local development (with A365 observability)

Copy .env.template to .env and fill in all values from a365 setup all. The template sets AGENT365_USE_MANAGED_IDENTITY=false, which tells the token service to authenticate using the blueprint's client secret instead of Managed Identity.

cd python/autonomous/github-trending
cp .env.template .env   # then fill in the values from a365 setup all
pip install -e .         # or: uv pip install -e .
python main.py

The agent starts on http://localhost:3979. Console output shows:

Observability token registration on startup
Trending digest immediately, then every 60 seconds
Heartbeat log every 60 seconds

Deploying the Agent

Deploy to your hosting provider and ensure:

Managed Identity is enabled on the hosting resource — the MSI must have a Federated Identity Credential (FIC) configured against the blueprint app
All AGENT365_* environment variables are set (except AGENT365_CLIENT_SECRET, which is only needed if MSI is unavailable)
AGENT365_USE_MANAGED_IDENTITY can be omitted — it defaults to true, so the token service uses MSI automatically
ENABLE_A365_OBSERVABILITY=true and ENABLE_A365_OBSERVABILITY_EXPORTER=true must be set for observability spans and export
The 3-hop FMI chain in production: MSI → Blueprint FIC (fmi_path) → Agent Identity → Observability API token

Observability Architecture

Tracing spans

Each autonomous cycle produces three nested spans:

InvokeAgentScope                        (root — wraps the entire cycle)
  |-- InferenceScope                    (child — wraps the LLM call)
  |-- ExecuteToolScope                  (child — wraps the GitHub API tool call)

Token flow — 3-hop FMI chain

The observability_token_service.py background task acquires tokens using:

Local dev (client secret):
  Blueprint ConfidentialClient + ClientSecret + fmi_path(agentId)
      -> T1 token (targeted at Agent Identity)
      -> Agent Identity ConfidentialClient + T1 as assertion
      -> AcquireTokenForClient("api://9b975845-.../.default")
      -> Observability API token

Production (MSI):
  MSI -> ManagedIdentityCredential.get_token("api://AzureADTokenExchange")
      -> Blueprint ConfidentialClient + assertion + fmi_path(agentId)
      -> T1 token (targeted at Agent Identity)
      -> Agent Identity ConfidentialClient + T1 as assertion
      -> Observability API token

The token is refreshed every 50 minutes and cached in token_cache.py.

For more details, see Agent observability — Microsoft Learn.

Support

For issues, questions, or feedback:

Issues: Please file issues in the GitHub Issues section
Documentation: See the Microsoft Agent 365 Developer Documentation
Security: For security issues, please see SECURITY.md

Contributing

This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.opensource.microsoft.com.

When you submit a pull request, a CLA bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., status check, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact opencode@microsoft.com with any additional questions or comments.

Additional Resources

Trademarks

Microsoft, Windows, Microsoft Azure and/or other Microsoft products and services referenced in the documentation may be either trademarks or registered trademarks of Microsoft in the United States and/or other countries. The licenses for this project do not grant you rights to use any Microsoft names, logos, or trademarks. Microsoft's general trademark guidelines can be found at http://go.microsoft.com/fwlink/?LinkID=254653.

License

Licensed under the MIT License — see the LICENSE file for details.