Teal Agent Framework is a (early!) prototype framework meant to accelerate the creation and deployment of AI-powered agents. The framework is built on top of Microsoft's Semantic Kernel, a lightweight, open source, agent framework.
Building upon Semantic Kernel, the Teal Agent Framework takes a config-first approach to the creation of agents. The majority of setup is performed in an agent's configuration file and you have the option to add in custom code to enhance your agent's capability.
A2A (Agent-to-Agent) Functionality Deprecated: As part of ongoing framework migration evaluation, A2A functionality in AppV2 (skagents/v2alpha1) is being deprecated. The A2A feature set was experimental and never fully fleshed out or stabilized for production use. While existing A2A implementations will continue to function for backward compatibility, new development should avoid using A2A functionality. The A2A feature set is no longer actively developed or enhanced.
- Python 3.11 or higher
- An appropriate API key for the LLM of your choice
- Docker (or comparable equivalent)
Running locally will allow you to test your agent's configuration and code.
First, clone the repository locally and install all dependencies. In this
I'm using uv as an environment manager.
$ git clone https://github.com/MSDLLCpapers/teal-agents
$ cd teal-agents/src/sk-agents
$ uv syncOnce cloned, you'll need to set up an environment file which will provide your
LLM API key and point to the correct agent configuration file. Create a .env
file in the root of the repository and add the following:
TA_API_KEY=<your-API-key>
TA_SERVICE_CONFIG=demos/01_getting_started/config.yaml
Finally, start the agent using fastapi via either using uv run or after
activating your environment.
$ uv run -- fastapi run src/sk_agents/app.pyor
$ source .venv/bin/activate
$ fastapi run src/sk_agents/app.pyYou can test the agent by visiting http://localhost:8000/docs
