Skip to content
Get packs

Prompt flow

Open-source toolkit for building, testing, evaluating and deploying LLM apps, from prototype to production.

GitHub Download promptflow_1.17.1
Stars
11.3k
Forks
1.1k
Licence
MIT
Language
Python
Last update
6 days ago
Type
Dev tool

What it is

Prompt flow is a set of development tools for building applications that use large language models (LLMs). You connect LLMs, prompts and Python code into a flow, then debug it, run it against larger datasets to evaluate quality, and deploy it. The README notes that Prompt flow is being retired by Microsoft on April 20, 2027, and is no longer recommended for new development.

  • Create a chatbot flow from a template with the pf flow init command
  • Connect an OpenAI or Azure OpenAI key and test a flow interactively
  • Run batch tests and evaluations on a flow with a larger dataset
  • Follow the Chat with PDF tutorial to build a question-answering app with evaluation
  • Migrate an existing flow using the migration guide and code samples

Great for

  • Teams with existing Prompt flow applications that need to keep running until migration
  • Developers who want to trace and evaluate LLM app behavior on datasets
  • Engineers who want to run flow tests in CI/CD pipelines
  • Developers who prefer a VS Code extension for visual flow editing

Not ideal for

  • Anyone starting a new project: the README says Prompt flow is no longer recommended for new development and suggests Microsoft Agent Framework instead
  • Teams relying on Prompt flow runtime container images, which no longer receive security or package updates
  • Non-developers, since setup uses the command line and a Python environment
Step-by-step guide: how to use Prompt flowInstall, first run, examples, tips and FAQs
Prompt flow alternativesPrompt flow vs promptfooBotpress vs Prompt flowLangChain vs Prompt flowLlamaIndex vs Prompt flowDSPy vs Prompt flow

How to get it

  1. 1

    Install with a package manager

    You'll need: Python 3.9 to 3.11 (recommended by the README)

    Install the packages

    pip install promptflow promptflow-tools
  2. 2

    Run it

    • Run pf flow init --flow ./my_chatbot --type chat to create a chatbot flow
    • Create a connection with pf connection create, using your API key
    • Run pf flow test --flow ./my_chatbot --interactive to chat with the flow
    • Follow the 15-minute tutorial for prompt tuning, batch testing and evaluation
  3. 3

    Get the most out of it

    • Use the VS Code extension for an interactive flow designer
    • Use tracing to inspect LLM interactions while debugging flows
    • Run evaluations on larger datasets before deploying
    • Integrate flow testing and evaluation into your CI/CD system
    • Review the migration guide if you plan to move to Microsoft Agent Framework

Downloads

Source code (.zip)main branch · straight from GitHubRelease promptflow_1.17.11 files incl. promptflow-1.17.1.tar.gzOfficial website & docsmicrosoft.github.io/promptflow/

Strengths

  • Covers the full cycle from prototyping through testing, evaluation and deployment in one toolkit
  • Includes a VS Code extension and a command-line interface for flow development
  • Released under the MIT license with an active public repository

Trade-offs

  • The README states retirement on April 20, 2027, and no new-development recommendation, so long-term support is limited
  • Container images are no longer updated, which is a security concern for deployed workloads
  • Setup and tutorials assume familiarity with Python and LLM API keys

How it compares

ProjectStarsTypeLevelUpdatedBest for
promptflow (this)11.3kDev toolintermediate6 days agoTeams with existing Prompt flow applications that need to keep running until migration
langchain149kFrameworkintermediate4 days agoPython developers building LLM-powered applications
llama_index52.8kLibraryintermediate10 days agoPython developers building retrieval-augmented generation (RAG) apps over private documents
dspy38.9kLibraryintermediate5 days agoPython developers building LM-based pipelines such as RAG or multi-step agents
botpress15kDev toolintermediate6 days agoDevelopers who want to write integrations and bots as TypeScript code using the Botpress CLI and SDK
generative-ai-for-beginners123kLearning resourcebeginner5 days agoDevelopers who know some Python or TypeScript and want a guided introduction to generative AI

Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.