Integrations / Microsoft Agent Framework
Add Badgr to Microsoft Agent Framework agents
Use Badgr as the model for your agents through an OpenAI-compatible client, and expose Badgr jobs as a function tool.
A setup guide only. Badgr has not tested this path, so expect to adjust it.
Why use Badgr for Microsoft Agent Framework
- Badgr interface
- Run
- Where Microsoft Agent Framework fits
- External framework
- Why Badgr
- Enterprise agent compute
- Use Badgr as the model through an OpenAI-compatible client, and as a tool for heavy steps.
- Every job has a maximum cost set in your tool code.
- Job events can be read through the executions API, with at-least-once delivery and stable event ids.
What you can do
- Enterprise agents that call a hosted open model
- A function tool that runs a heavy step as a capped job
- Check on a job later from a separate workflow step
Quickstart: Integration outline
- Create the framework's OpenAI-compatible chat client with base URL https://aibadgr.com/v1, your Badgr key and the model badgr-auto. Check the framework's documentation for the current client class.
- Define a function tool that starts a Badgr job with a maximum cost, using the CLI or the jobs API.
- Return a job id from the tool and add a second tool that reads its status and result.
How it works
- 1.Your agent calls Badgr for model replies
- 2.A tool starts a capped job for heavy work
- 3.A second tool reads the job result
- 4.The agent continues with the output
What you get
- Agent replies from Badgr models
- Job results returned through your tools
Requirements
- A Badgr API key (get one at /get-api-key)
- An existing Microsoft Agent Framework project
- For jobs: the Badgr CLI or the jobs API and a funded balance
What it costs
- Model calls are billed per request. Jobs are billed for run time and bounded by a maximum cost you set in the tool.
- Badgr does not provide callbacks into your agent. Poll for status instead.
See pricing for how billing works.
Proof
No tested run is recorded for this integration yet. A receipt and example output will be added here after a real run, not before.
Troubleshooting
Client class or options differ
The framework changes quickly. Follow its documentation for the OpenAI-compatible client and only change the base URL, key and model.
Task callbacks expected
Poll the job status. Job events are also available through the executions API with at least once delivery.
Request shape unclear
The jobs API is documented in the compute API reference. Start there for the exact request shape and base URL.
Developer reference
Ready to start with Microsoft Agent Framework?
Every run or service has a spending limit you set before it starts.