Open-source LLM observability and evaluation platform for tracing AI agents, scoring LLM outputs, managing prompts, and monitoring production.
microsoft/promptflow
A suite of developer tools for building, testing, evaluating and deploying LLM-based applications, with a VS Code flow designer and a pf CLI.
- Stars
- 11,254
- Forks
- 1,127
- Open issues
- 69
- Last commit
- 2 days ago
What is it?
Prompt flow covers the development cycle of LLM-based AI applications, from prototyping and testing through evaluation to production deployment and monitoring. Developers link LLMs, prompts, Python code and other tools into executable flows, then debug them with tracing of LLM interactions.
Flows can be run against larger datasets to measure quality and performance, and the testing and evaluation steps can be wired into a CI/CD pipeline. Flows can be deployed to a serving platform of choice or embedded directly into an application's codebase.
The project provides a pf command-line interface, a VS Code extension with a visual flow designer, and an optional cloud version in Azure AI. Note that Microsoft has announced that Prompt flow in Microsoft Foundry and Azure Machine Learning will be retired on April 20, 2027, and is no longer recommended for new development.
Key features
- 01Build executable flows that combine LLMs, prompts, Python code and tools.
- 02Trace and debug LLM interactions within a flow.
- 03Run flows over large datasets to evaluate quality and performance.
- 04Integrate testing and evaluation into CI/CD pipelines.
- 05Use a pf CLI and a VS Code extension with a visual flow designer.
Use cases
- Prototyping a chatbot or retrieval-based question answering app and iterating on its prompts.
- Batch-testing prompt changes against a dataset before shipping them.
- Adding automated LLM quality checks to a team's CI pipeline.
Quick start
- 1
Install the packages with pip
pip install promptflow promptflow-tools - 2
Create a chatbot flow from the chat template
pf flow init --flow ./my_chatbot --type chat - 3
Set up a connection for your OpenAI API key
pf connection create --file ./my_chatbot/openai.yaml --set api_key=<your_api_key> --name open_ai_connection - 4
Chat with your flow interactively
pf flow test --flow ./my_chatbot --interactive
Things to know
- The README warns that Prompt flow will be retired on April 20, 2027 and is no longer recommended for new development.
- Prompt flow container images no longer receive security or package updates.
- Running the chatbot quick start requires an OpenAI or Azure OpenAI API key.
Topics
FAQ
- What is promptflow?
- A suite of developer tools for building, testing, evaluating and deploying LLM-based applications, with a VS Code flow designer and a pf CLI. Prompt flow covers the development cycle of LLM-based AI applications, from prototyping and testing through evaluation to production deployment and monitoring. Developers link LLMs, prompts, Python code and other tools into executable flows, then debug them with tracing of LLM interactions.
- Is promptflow free to use? What license does it have?
- promptflow is open source under the MIT license. Check the license terms before using it in commercial products.
- Is promptflow still maintained?
- Actively maintained. The last commit to promptflow was on Oct 20, 2026.
- What can I use promptflow for?
- Prototyping a chatbot or retrieval-based question answering app and iterating on its prompts. Batch-testing prompt changes against a dataset before shipping them. Adding automated LLM quality checks to a team's CI pipeline.
- How do I install promptflow?
- Install the packages with pip: pip install promptflow promptflow-tools Create a chatbot flow from the chat template: pf flow init --flow ./my_chatbot --type chat Set up a connection for your OpenAI API key: pf connection create --file ./my_chatbot/openai.yaml --set api_key=<your_api_key> --name open_ai_connection Chat with your flow interactively: pf flow test --flow ./my_chatbot --interactive
- Where can I get promptflow?
- The source code and installation instructions are on GitHub: https://github.com/microsoft/promptflow
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Python framework for building LLM applications with multiple cooperating agents, tools, and retrieval-augmented generation.
Open-source platform for tracing, testing, routing and governing LLM calls, agents and coding assistants across a company.
A community-curated open-source library of prompts for AI chat models, with a website, CSV/Hugging Face datasets, CLI and MCP server integration.
A Python framework that benchmarks LLMs by having them play the Bulls and Cows code-breaking game through LiteLLM-compatible providers.
Jan is an open-source desktop app for downloading and running LLMs locally, offering a ChatGPT-like interface with an optional cloud model connection.