# Python

Transparent AI agent framework with 100+ skills and real-time visibility

Agent Zero is an open-source general-purpose AI agent framework with 16,700+ GitHub stars that uses the computer itself as a tool. Unlike structured orchestration frameworks, it provides full transparency where every thought, action, and tool call is visible and editable in real time, supporting 100+ extensible skills.

Browser-based autonomous AI agent platform

AgentGPT is an open-source browser-based platform with 36K+ GitHub stars for creating and deploying autonomous AI agents without any setup or installation. Give an agent a name and goal in plain language, and it autonomously decomposes the objective into subtasks, executes them, and iterates toward completion. Built with Next.js and supports multiple LLM providers. Features include web search, code execution, and task chaining. No technical expertise required to create functional AI agents.

Production-ready multi-agent platform by Alibaba

AgentScope is an open-source multi-agent platform with 22K+ GitHub stars developed by Alibaba. Designed for production-ready multi-agent applications with built-in distributed execution, fault tolerance, and agent-to-agent messaging. Features memory management with compression, a drag-and-drop workstation for visual agent building, multi-modal support, and flexible pipelines for sequential, parallel, and conditional agent orchestration. Supports all major LLM providers.

Security scanner for AI agentic workflows and MCP servers

Agentic Radar is an open-source CLI security scanner that maps attack surfaces in agentic AI workflows. It detects MCP servers, visualizes agent tool chains, and validates against OWASP LLM Top 10 vulnerabilities including prompt injection and excessive agency. Supports scanning CrewAI, LangGraph, AutoGen, and Semantic Kernel pipelines. Built by SPLX AI with active development and MCP-specific detection capabilities added for the growing MCP ecosystem.

Non-agent approach to automated software engineering via localize-and-repair

Agentless takes a deliberate non-agent approach to LLM-powered software engineering. Instead of autonomous agents making tool calls, it uses a structured localize-then-repair pipeline: first narrowing down which files and functions are relevant, then generating targeted patches. Achieved competitive SWE-Bench results at $0.34 average cost per issue. Adopted by OpenAI for o3 evaluations. 3,000+ GitHub stars, MIT licensed. A counterpoint to the agent-heavy trend in AI coding tools.

ELT platform with 350+ data connectors

Airbyte is an ELT platform with 350+ pre-built connectors for syncing data from any source to warehouses, lakes, and AI pipelines. It handles incremental syncs, schema evolution, and change data capture with a connector builder for custom integrations. Used by DoorDash, Replit, and thousands of data teams. Over 15,000 GitHub stars and $150M+ in funding.

Context retrieval layer for AI agents and RAG

Airweave is an open-source context retrieval platform that connects AI agents and RAG systems to 50+ apps and databases through a unified search interface. It continuously syncs data from sources like Notion, Slack, GitHub, and databases, making it searchable through LLM-friendly APIs. Airweave includes Python and TypeScript SDKs, MCP support, and a CLI for managing data connections.

AWS managed AI code review for Java and Python

Amazon CodeGuru Reviewer is an AWS managed service that provides automated code analysis and recommendations on pull requests for Java and Python codebases. It integrates with GitHub, Bitbucket, and AWS CodeCommit, using machine learning trained on Amazon's internal code review practices to identify bugs, security vulnerabilities, and performance issues with IAM-based access control for enterprise governance.

Official SDK for building Claude-powered agentic applications

Anthropic's official SDK for building agents with Claude. Provides high-level abstractions for tool use, multi-turn conversations, computer use, and agent loops on top of the Claude API. Simplifies the development of production-grade agents by handling common patterns like retry logic, context management, and tool orchestration in a well-tested library.

Workflow orchestration platform for data pipelines

Apache Airflow is an open-source workflow orchestration platform with 39K+ GitHub stars for authoring, scheduling, and monitoring data pipelines as Python DAGs. Used by 80K+ organizations for ETL, ML training, and data transformation. Features dynamic pipeline generation, extensive operator library for AWS/GCP/Azure, task dependencies, retries, SLA monitoring, a rich web UI with Gantt charts, and pluggable executors from local to Kubernetes. The industry standard for pipeline orchestration.

Tool platform for AI agents

Arcade AI is an MCP runtime platform for secure agent authorization, reliable tool execution, and governance. It enables enterprises to deploy authenticated AI agents across Gmail, Slack, Salesforce, and 100+ other services without managing OAuth complexity or token refresh cycles. Arcade provides a centralized authorization layer that handles authentication, permission scoping, and compliance for every tool call an agent makes.

Open-source data curation platform for LLM fine-tuning

Argilla is an open-source platform for curating and annotating data for LLM fine-tuning and RLHF workflows. It provides collaborative annotation interfaces for text classification, ranking, and preference labeling with integrated quality metrics. Part of the Hugging Face ecosystem, Argilla supports direct dataset publishing to the Hub and integrates with major training frameworks for seamless model improvement pipelines.

Build AI agents like LEGO — modular and predictable

Atomic Agents is a lightweight Python framework by BrainBlend AI that applies Atomic Design principles to AI agent development. Each component — agents, tools, context providers — is a single-purpose, reusable building block with Pydantic-enforced input/output schemas for type safety. Built on Instructor for structured LLM outputs, it prioritizes predictability and developer control over the autonomous-but-unpredictable behavior of larger frameworks like LangChain or CrewAI.

Open-source autonomous AI agent platform

AutoGPT is an open-source autonomous AI agent platform with 183K+ GitHub stars that breaks goals into subtasks and executes them independently. Features a visual Agent Builder for creating workflows without coding, persistent cloud-based agents running on triggers, a marketplace of pre-built agents, and a plugin system. Agents can browse the web, write code, manage files, and call tools autonomously while maintaining memory across sessions.

Type-safe LLM function builder

BAML is a domain-specific language by BoundaryML for building reliable AI workflows and agents through schema engineering. It turns prompt engineering into a structured, type-safe discipline by letting developers declaratively define function schemas, validate LLM responses, and version prompts without fragile JSON parsing or boilerplate. BAML reframes prompt engineering as schema definition, making AI workflows testable and maintainable across models.

Python and TypeScript framework for production multi-agent systems

BeeAI Framework is an Apache-2.0 toolkit for building production-ready AI agents and multi-agent systems in Python and TypeScript. Its docs cover agents, tools, RAG, memory, workflows, backend providers, serving, and A2A/MCP integration surfaces, making it a vendor-neutral option for teams comparing LangGraph, CrewAI, Mastra, and related agent runtimes.

State machine framework for AI agent tracking

Burr is an open-source Python framework for building applications as state machines with built-in observability and persistence. Designed for AI agents, chatbots, and RAG pipelines where tracking state transitions, debugging decisions, and replaying execution is critical. Features automatic state persistence, a visual UI for inspecting execution flows, time-travel debugging, and integration with LangChain, OpenAI, and any Python code. Apache 2.0 licensed by DAGWorks.

Multi-agent software company simulation for automated development

ChatDev simulates an entire virtual software company through multi-agent collaboration where LLM-powered roles including CEO, CTO, programmer, tester, and designer work together to produce complete software. With 32,000+ GitHub stars and a NeurIPS 2025 accepted paper, it offers a novel approach to automated software development through role-based agent orchestration.

State-of-the-art open-source text-to-speech with emotion control

Chatterbox is an open-source text-to-speech model by Resemble AI that delivers state-of-the-art voice synthesis with fine-grained emotion and style control. The model supports zero-shot voice cloning from short audio samples, produces natural-sounding speech across multiple speaking styles, and runs locally without cloud dependencies. With over 24,000 GitHub stars, it has become the leading open-source alternative to commercial TTS services for developers building voice-enabled AI applications.

Official agent SDK by Anthropic

Anthropic's Python SDK for building agentic AI applications powered by Claude models. Provides primitives for creating agents with tool use, multi-step reasoning, guardrails, handoffs between specialized agents, and structured output. Supports building complex agent workflows with tracing and observability. Designed for developers building production AI agents that interact with external systems, databases, and APIs using Claude as the reasoning backbone.

Self-hosted cloud development environments for teams and AI agents

Coder provisions self-hosted cloud development environments on any infrastructure including Kubernetes, Docker, AWS, GCP, and Azure. Developers connect through VS Code, JetBrains IDEs, or browser-based editors to standardized environments with pre-configured dependencies. Features template-based provisioning, automatic shutdown, and audit logging. Over 12,800 GitHub stars with growing AI agent use cases.

Workflow orchestration engine

Netflix-originated workflow orchestration platform with JSON and code-based workflow definitions, human-in-the-loop support, and AI agent orchestration capabilities. With 18k+ GitHub stars, Conductor handles complex distributed workflows at massive scale, offering built-in retry logic, event-driven triggers, and visual workflow monitoring for microservice coordination.

Open-source deep learning text-to-speech toolkit

Coqui TTS is an open-source deep learning toolkit for text-to-speech synthesis, originally built by former Mozilla TTS engineers. It supports multi-speaker and multilingual synthesis, voice cloning from just six seconds of audio, and ships pre-trained models for 20+ languages. After Coqui shut down in 2023, the Idiap Research Institute forked and actively maintains it. With 45K+ GitHub stars, it remains the most popular open-source TTS framework in Python.

Production-grade web scraping and browser automation library

Crawlee is an open-source web scraping and browser automation library for Node.js and Python that handles the hard parts of building reliable crawlers. It manages proxy rotation, request queuing, automatic retries, session management, and fingerprint spoofing out of the box. Supports Puppeteer, Playwright, Cheerio, and HTTP-based crawling with a unified API. Built by Apify, it includes persistent storage, autoscaling concurrency, and TypeScript-first design for production deployments.

Programming — not prompting — LLMs

Declarative framework from Stanford University for programming language models rather than prompting them. DSPy treats LLM interactions as programmable modules with input-output signatures and uses optimization algorithms to automatically compile these modules into effective prompts or fine-tuned weights, replacing brittle prompt strings with structured, modular AI software.

3D reconstruction without camera parameters

DUSt3R is Naver's breakthrough 3D reconstruction method that generates dense 3D scenes from unconstrained image pairs without known camera intrinsics or extrinsics. It casts pairwise reconstruction as pointmap regression, removing hard geometric constraints of projective camera models. Supports multi-view alignment, depth estimation, visual localization, and extends to MASt3R and MUSt3R for large-scale applications.

Git-based version control for ML data and pipelines

DVC (Data Version Control) is a free open-source tool that brings Git-like version control to datasets, ML models, and experiment pipelines. It stores pointer files in Git while keeping large data in remote storage like S3, GCS, or Azure. Features include reproducible ML pipelines with DAG-based dependency tracking, experiment management, metrics comparison, and a VS Code extension for visual experiment tracking.

High-performance data engine for multimodal AI workloads

Daft is a high-performance distributed data engine designed specifically for AI and multimodal workloads. It processes structured data alongside images, audio, video, and embeddings natively, outperforming Spark and Polars on AI-specific data pipelines. Built in Rust with a Python API, Daft handles the data engineering challenges unique to machine learning workflows.

DeepSeek's expert-parallel communication library for MoE model training

DeepEP is DeepSeek's open-source communication library optimized for expert-parallel training of Mixture-of-Experts models. It provides efficient GPU-to-GPU data routing for distributing tokens to expert networks across multiple devices during MoE model training and inference. Enables the distributed expert parallelism that powers DeepSeek's competitive model efficiency. Over 9,100 GitHub stars.

DeepSeek's FP8 general matrix multiplication kernels for efficient inference

DeepGEMM is DeepSeek's open-source library of FP8 matrix multiplication CUDA kernels optimized for LLM inference and training on modern NVIDIA GPUs. It provides efficient GEMM operations using 8-bit floating point precision that reduce memory bandwidth requirements while maintaining model accuracy. Designed for integration into inference engines and training frameworks. Over 6,300 GitHub stars.

State-of-the-art open-source code language models

DeepSeek Coder is a family of open-source code language models trained from scratch on 2 trillion tokens of code and natural language data. Available in sizes from 1B to 33B parameters, these models support 80+ programming languages with 16K context windows and fill-in-the-blank capabilities. DeepSeek Coder outperforms CodeLlama-34B on HumanEval and MBPP benchmarks while being commercially licensable under MIT.

Deep learning optimization for distributed training

DeepSpeed is Microsoft's open-source deep learning optimization library that makes distributed training and inference easy, efficient, and effective. Its ZeRO optimizer eliminates memory redundancies across data-parallel processes, enabling training of models with trillions of parameters. DeepSpeed supports 3D parallelism combining data, pipeline, and tensor parallelism, along with mixed precision training, gradient checkpointing, and CPU/NVMe offloading for memory-constrained environments.

Open-source LLM red-teaming framework with 40+ attack types

DeepTeam is an open-source red-teaming framework for systematically testing LLM applications against 40+ adversarial attack types. It covers OWASP Top 10 for LLMs including jailbreaks, prompt injection, PII leakage, and hallucination attacks. Built as the sister project of DeepEval for security testing alongside evaluation. Apache-2.0 licensed.

ByteDance's open-source agentic workflow platform

DeerFlow is ByteDance's open-source agentic workflow platform that combines LLM-powered agents with deep research capabilities. It features multi-agent orchestration, web research automation, report generation, and podcast creation from research outputs. Built with LangGraph, it supports configurable LLM backends and provides a modern React-based UI for managing complex AI-driven research and content workflows.

Get your documents ready for gen AI

Docling is an open-source document processing toolkit by IBM Research that converts complex documents into structured formats optimized for generative AI applications. It parses PDF, DOCX, PPTX, XLSX, HTML, images, audio, and LaTeX with advanced PDF understanding including layout analysis, reading order detection, and table structure recognition. Docling exports to Markdown, HTML, JSON, and DocTags, and integrates natively with LangChain, LlamaIndex, and other AI frameworks for RAG workflows.

AI-native SAST with contextual security analysis

DryRun Security is an AI-native SAST platform using Contextual Security Analysis to reason about code behavior, data flow, and exploitability instead of regex pattern matching. It provides PR-native security reviews on GitHub and GitLab, catching logic flaws, broken auth, IDOR, and injection bugs that legacy scanners miss while cutting 90% of noise. Features Natural Language Code Policies, DeepScan for full-repo audits, and a Risk Register for org-wide visibility. Supports 14+ languages.

Postgres sync engine for local-first and real-time applications

ElectricSQL is a sync engine that keeps local application state synchronized with PostgreSQL in real-time. It enables local-first architectures where apps work offline with instant responsiveness, syncing data bidirectionally when connectivity is available. Supports partial replication with shape-based subscriptions to sync only relevant data subsets to each client.

Prompt engineering framework treating prompts as versioned Python functions

Ell is a prompt engineering library that treats LLM prompts as versioned, testable Python functions rather than opaque strings. Built by ex-OpenAI researcher William Guss, it provides automatic prompt versioning with content-addressable hashing, a local TensorBoard-like studio for visualizing prompt evolution, and structured output support via Pydantic. 5,800+ GitHub stars, MIT licensed. Designed for teams who want to version-control and systematically improve their prompts over time.

Multimodal data labeling and curation for production AI

Encord is a data labeling and curation platform for teams building production AI systems with complex multimodal data. It supports image, video, audio, DICOM medical imaging, and 3D point cloud annotation with AI-assisted labeling, advanced ontology management, and quality assurance workflows. Features active learning for prioritizing high-value samples and integrates with major ML frameworks.

PyTorch on-device AI for mobile and edge devices

ExecuTorch is PyTorch's official solution for deploying AI models on mobile, embedded, and edge devices. It features a 50KB base runtime, 12+ hardware backends including Apple CoreML, Qualcomm QNN, ARM, and Vulkan, and native PyTorch export without format conversions. Powers Meta's on-device AI across Instagram, WhatsApp, Quest 3, and Ray-Ban Smart Glasses, supporting LLMs, vision, speech, and multimodal models.

Python toolkit for assessing and mitigating ML model fairness issues

Fairlearn is a Microsoft-backed open-source Python toolkit that helps developers assess and improve the fairness of machine learning models. It provides metrics for measuring disparity across groups defined by sensitive features, mitigation algorithms that reduce unfairness while maintaining model performance, and an interactive visualization dashboard for exploring fairness-accuracy trade-offs. Integrated with scikit-learn and Azure ML's Responsible AI dashboard.

Zero-config conversion of FastAPI endpoints into MCP tools

FastAPI-MCP automatically converts existing FastAPI endpoints into Model Context Protocol tools with zero configuration. With 11,700+ GitHub stars, it is the most popular MCP integration layer, enabling thousands of FastAPI developers to make their APIs accessible to AI agents like Cursor and Claude through the MCP standard.

Python web framework with HTMX, no JavaScript needed

FastHTML is an open-source Python web framework by Answer.AI that renders HTML directly from Python functions using HTMX. It eliminates the JavaScript build step entirely, letting developers create interactive web apps with pure Python. FastHTML is designed for AI demo apps, admin tools, and internal tooling where full-stack simplicity matters more than SPA complexity. Built by Jeremy Howard's team.

Open-source feature store for machine learning

Feast is an open-source feature store that manages and serves ML features for both training and online inference. It prevents training-serving skew by providing consistent feature access across offline and real-time environments. Feast supports batch materialization from data warehouses, real-time feature retrieval, on-demand transformations, and integrates with major data platforms including BigQuery, Snowflake, Redshift, and DynamoDB.

Generate SDKs and API documentation from OpenAPI specs

Fern is a platform that generates idiomatic client SDKs and beautiful API documentation from OpenAPI specifications. It produces type-safe SDKs for TypeScript, Python, Java, Go, Ruby, and C# with features like pagination, retries, and streaming built in. The documentation includes API reference, guides, and an AI-powered search assistant. Used by companies like Cohere, ElevenLabs, and Merge.

Multilingual emotional text-to-speech with 80+ language support

Fish Speech is an open-source text-to-speech system supporting 80+ languages with emotional expression, zero-shot voice cloning, and real-time streaming. It generates natural speech with controllable emotions, speaking styles, and prosody. Features a web interface, API server, and integration with AI agent frameworks for voice-enabled applications. Over 29,000 GitHub stars.

Open-source feature flags and remote configuration

Flagsmith is an open-source feature flag and remote config platform with 15+ SDKs, a clean REST API, and simple Docker deployment. It supports boolean, multivariate, and string flags with segment-based targeting, A/B testing via flag variations, and environment-based workflows. Flagsmith provides a straightforward self-hosted alternative to LaunchDarkly with full API-first design under a BSD-3-Clause license.

Fast memory-efficient GPU attention kernels

FlashAttention is a fast and memory-efficient exact attention implementation that reduces GPU memory usage from quadratic to linear in sequence length. Created by Tri Dao, it achieves 3-4x speedups over baseline implementations through IO-aware tiling that minimizes HBM reads and writes. Versions include FlashAttention-2 with improved parallelism, FlashAttention-3 optimized for Hopper H100 GPUs, and FlashAttention-4 targeting Hopper and Blackwell architectures.