Awesome AITools
Collection of AI-related utilities. Welcome to submit pull requests /收藏AI相关的实用工具，欢迎提交pull requests
Awesome AI Tools
English | 中文
This repo collects awesome AI tools. Welcome everyone to recommend more awesome AI tools together! Please use the following template as a reference for your recommendations. Contribution guidelines
💎 Sponsor
A huge thank you to our sponsors! Want to appear here? Contact us.
Click to collapse
 Xuanshu API | Xuanshu API — a next-generation AI model routing gateway for enterprises, technical teams, and individual developers, providing one-stop API access to world-class top models (Claude, GPT, Grok, etc.) with enterprise-grade stability. Recharge and enjoy 20% off, models starting from 20% of the original price, $5 free upon registration, invoice support for enterprises. Click this link to register and get an extra $5 credit. |
 Infistar.cc | Infistar.cc — a one-stop global LLM API platform. One key works with every tool in this list: plug & play, any tool that supports a custom Base URL connects directly, with one shared balance across tools; official API access with full-power models and stable multi-route balancing, plus stable direct connectivity in China and CNY settlement; 18+ providers and 190+ models (Claude / GPT / DeepSeek, etc.) — switching models is just changing the name; select models from as low as 1% of official pricing, plus account top-up service for GPT / Claude / Gemini. Sign up via this exclusive link to get a $5 trial credit. |
All Categories
- AI Assistants & Proprietary Models
- Decision Models
- Open Source LLMs
- LLM Leaderboard
- AI Agent
- Coding Agents
- Code Review & Development Utilities
- Agent Workspaces & Collaboration
- Agent Frameworks & Runtimes
- Browser & Computer Automation
- Agent Memory & Context
- Agent Skills
- AI News & Information
- AI Security & Reverse Engineering
- General LLM Applications
- Office Collaboration CLI / MCP
- AI Finance & Quant Investment
- AI Image Creation and UI Design
- Video Creation
- AI Infrastructure Platform
- GPU Programming
- LLM Prompts
- LLM training platform
- LLM Inference & Deployment
- Writing
- Translation
- Speech Recognition & Subtitles
- Text to Speech (TTS)
- Voice Cloning & Conversion
- Speech Processing & Translation
- AI Music & Sound Effects
- Video Content Summary
- Academic research
- OCR
- AI Detection
- Humanoid Robots
- Embodied AI & Simulation
AI Assistants & Proprietary Models
| Name | Description | Links | Fees |
|---|---|---|---|
| ChatGPT | 🌟 OpenAI's AI assistant. Current flagship: GPT-6 Astra; GPT-6.1 Sol from the Sep 29 DevDay and GPT-6 Luna cut API prices ~50% ($2/$10 and from $0.10 input per M tokens) and are free within consumer limits. Starting Oct 7, 2026, GPT-6 (Sol/Luna) with Intelligent UI is rolling out globally to all ChatGPT users: paid tiers default to GPT-6 Sol, Free/Go tiers get GPT-6 Luna, and responses can automatically render interactive charts, forms, and visual layouts. Differentiation: persistent memory and user profile — remembers context across conversations and feels the most personalized; best general-purpose assistant for coding, writing, and daily tasks. More Info | 1. URL 2. API | Free/Paid |
| Claude | 🌟 Anthropic's AI assistant. Current flagship: Claude Fable 5.1 (general) / Claude Mythos 5.1 (trusted access); Claude Opus 5.5 matches Fable 5.1 on most tasks at ~40% lower cost, Sonnet 5.5 shipped Sep 28, and Haiku 5.5 launched Oct 7 at $0.10/$0.50 per M tokens for prompts up to 100k — roughly 1/10 the price of Haiku 4.5 and 1/20 of Sonnet 5.5. Sonnet 5.5 cache reads were also cut 50%, making most agentic work ~20% cheaper. Differentiation: Cowork mode turns the chatbot into an agent that pulls financial data, builds Excel forecasting models, and executes workflows; strongest at coding, long-context reasoning, safety, and enterprise use. More Info | 1. URL 2. API | Free/Paid |
| Gemini | 🌟 Google's AI assistant combining native multimodality, Deep Research, and Google Workspace integration. Gemini 4 Argon is now available through the API, AI Studio, and Gemini Enterprise, with Gemini app access for Google AI Pro/Ultra subscribers; its 1M-token budget is shared by input and output. Gemini 3.8 Flash remains the fast option. Fairwind provides separate controlled access for trusted cyber defenders; Gemini Spark handles background tasks across Gmail, Calendar, and Tasks. More Info | 1. URL 2. API | Free/Paid |
| DeepSeek | DeepSeek's AI assistant, with DeepSeek-V4.1-Flash-1017 available on the web, app, and API. Focuses on reasoning, coding, and agent tasks at low API cost, with native multimodal understanding and MIT-licensed open weights. The October update improves complete-project development and reduces task runtime and output length; model pricing and the deepseek-flash API alias are unchanged. Benchmark gains are company-reported, and performance varies by task and harness. | 1. URL 2. API | Free/Paid |
| Kimi Chat | Moonshot AI's AI assistant. Current flagship: Kimi K3.1 Pro (Oct 24, 2026; first model of the K3.1 series) — a 3-trillion-parameter model with 1M-token context and major gains in coding agents, 3D creation, and finance workflows: Terminal-Bench 4.0 score 65.7% (vs GPT-6 Astra 58.2%), BenchCAD 92.7%, OfficeQA Pro v2 64.4 (current SOTA). K3.1 Pro Highspeed ships with the same intelligence at ~2-3x output speed; full model weights to be released by Nov 4, 2026. Differentiation: optimized for software engineering, repository-scale analysis, and long-document reasoning. More Info | 1. Kimi 2. Moonshot AI | Free |
| GLM | Zhipu AI's AI assistant. Current flagship: GLM-5.3; also offers GLM-5.3-Flash (native multimodal, MIT open weights, 1M context) with coding/agent performance near Claude Opus 4.8 at much lower cost. Differentiation: agentic coding, repo-scale analysis, cyber-defense, and long-context reasoning. | 1. URL 2. API | Free/Paid |
| Grok | SpaceXAI's AI assistant (formerly xAI). Current flagship: Grok 4.7 (Sept 21, 2026; focused on coding and knowledge work, 500K context, text+image input). Differentiation: real-time X data access — its moat is live internet/X posts as an information source; best for real-time news and events. | 1. URL 2. API | Free/Paid |
| Qwen | Alibaba's AI assistant. Current flagship: Qwen3.8-Max (GA Aug 3, 2026; open weights for Qwen3.8-27B released Aug 14). Differentiation: open-weight 2.4T MoE + long-horizon agentic reasoning — strong coding/deep research with 1M context and native multimodality. | 1. URL 2. API | Free |
| Dola | ByteDance's AI assistant. Current flagship: Doubao-Seed-2.1 Pro. Differentiation: clean, intuitive UI with strong general capabilities — straightforward chat experience for everyday tasks. | 1. URL 2. API | Free |
| Meta Muse | 🌱 Meta's personal AI agent for booking travel, shopping, forms, and email/calendar work. It runs tasks in an isolated VM and requests approval for sensitive actions. Muse Wonder is planned for subscriber-first rollout in Muse; Wonder is already available through the Meta Model API and Muse Code, while Wonder Full is offered to Muse Code subscribers. Regional rollout for adults, with free and paid plans; check the official site for current access. More Info | URL | Free/Paid |
Decision Models
| Name | Description | Links | Fees |
|---|---|---|---|
| Jev | 🌱 TypeSafe AI's "System One" decision model (early access) — it never generates text, returning typed answers (Choice/Score/Noul) with calibrated probabilities so confidence matches accuracy. Officially 20-200x faster and 1/40-1/400 cheaper than frontier LLMs on structured tasks; input $0.042/M tokens, output free, 70-500ms latency. Note: official figures, not yet independently verified. More Info | URL | Free/Paid |
| SemIf (formerly OpenJev) | 🌱 An MIT-licensed independent open-source project (explicitly unaffiliated with Jev or TypeSafe) that turns runtime semantic decisions into a probability readout: no training, no fine-tuning — it reads typed option probabilities directly from a frozen Qwen3.5-4B, generating no text and needing no JSON parsing, aimed at the small decisions agents make (route this, retry that, does the evidence hold?). On 21 yes/no judgments the direct readout has a median of 1.02s vs 5.33s for generating a JSON array token by token (~5.2x); agreement with TypeSafe's published 102-row subset is 84.5% (closed-source Jev reports 88.3%). Runs on a single RTX 3090, Apple Silicon (MLX/MPS), or CPU-only via llama.cpp, plus a browser-only WebGPU demo with no waitlist. | GitHub  Site/Demo | Free |
| OpenAI Decisions API | OpenAI's dedicated decision endpoint (POST /v1/decisions) in public beta. Returns typed answers — predicate (0–1 probability), choice (selected value + per-option probabilities), or score (probability-weighted mean across ordered levels) — based on gpt-6-luna. Input $0.10/M tokens; cache read/write and output tokens are free. About 10× faster than the same call through Responses API. | API Docs | Free/Paid |
| Microsoft-Decision-1 | Microsoft's fast decision-scoring model, post-trained from Qwen3.5-9B. Returns calibrated probabilities over fixed yes/no, multiple-choice, or rating options instead of generating text. Available through Microsoft Foundry and OpenRouter; 32K context, input $0.042/M tokens, output free. | Foundry OpenRouter | Free/Paid |
| Cloudflare Clef | Cloudflare Workers AI decision-model family. clef-omni (30B MoE / 3B active) accepts text, image, audio, and video in one request; clef-flash is optimized for low-latency scoring at $0.038/M input tokens. Context window up to 64K (24K for flash). | Workers AI Hugging Face | Free/Paid |
Open Source LLMs
| Name | Description | Links | Fees |
|---|---|---|---|
| Kimi K3 | 🌟 Moonshot AI's flagship open-weight model — the first open-source model in the 3-trillion-parameter class, putting near-frontier capability fully in the open. Modified MIT license; 1M-token context with native visual understanding, aimed at software engineering, knowledge work, and long-document reasoning. More Info | Technical Blog Hugging Face | Free |
| DeepSeek models | 🌟 DeepSeek's open-weight model family for reasoning, coding, and agent tasks. Current V4 models use the MIT license and support self-hosting and commercial use; V4.1-Flash-1017 offers native multimodal understanding, 1M-token context, and improved complete-project development and task efficiency, alongside V4-Pro. R1 (reasoning) and V3 (general-purpose MoE) remain available as earlier releases for research and existing deployments. More Info | V4 models Earlier releases: R1, V3 | Free |
| Qwen3 / Qwen3.8 | Alibaba Cloud's open-weight LLM/VLM families — the most complete open model lineup, strong in coding, multilingual, and agentic use. Qwen3.8 adds a dense vision-language model (image/video input) and Qwen3.8-Flash-Next, an ultra-cheap high-throughput variant previewing the Qwen4 architecture. | GitHub  Hugging Face | Free |
| GLM-5.3-Flash | Zhipu's open-weight (MIT) natively multimodal MoE model supporting 1M context — matches Claude Opus 4.8 on Zhipu's composite intelligence index at roughly 1/40 the API price. | Hugging Face | Free |
| Gemma 4 | Google's open-weight LLM series built on the Gemini architecture — Gemini-derived quality in small, deployable sizes (2B–31B) from on-device to server. Native vision/audio, up to 256K context, and strong multilingual support. | GitHub  | Free |
| Muse Glimmer | Meta's 30B-parameter open-weight multimodal model for local agents under Apache 2.0. Supports text/image input, tool use, and 4-bit quantization for consumer GPUs. Distilled from Muse Spark. | Hugging Face | Free |
| Muse Spark Open | Meta's open-weight multimodal model for coding and long-running agent tasks, with a 1M-token context. Runs locally through supported inference engines, including MLX on two 512GB M5 Ultra Mac Studios; substantial memory is still required. Uses the Muse Model License with limited commercial terms, not MIT or Apache-2.0: some commercial API resale requires a separate agreement. Hosted APIs are available from partner providers. | Hugging Face Official overview | Free weights / Paid API; license restrictions |
| Llama 4 | Meta's open-weight multimodal model family — the safest choice for ecosystem and deployment support. MoE variants range from Scout (10M-token context) to Maverick (400B total / 17B active); supports text and image input. | GitHub  | Free |
| Phi-4 | Microsoft's open-weight small language model (SLM) family — efficient small models for on-device and cloud deployment. Includes general, reasoning, multimodal, and vision variants (e.g., Phi-4-mini, Phi-4, Phi-4-multimodal, Phi-4-reasoning-vision). | GitHub  | Free |
LLM Leaderboard
| Name | Description | Links | Fees |
|---|---|---|---|
| LMArena (Chatbot Arena) | 🌟 LMArena (formerly LMSYS Chatbot Arena; renamed Arena in 2026) is a crowdsourced open platform for LLM evals, born from UC Berkeley's SkyLab (LMSYS, with Stanford/UCSD/CMU/MBZUAI). Collected over 1,000,000 human pairwise comparisons to rank LLMs with the Bradley-Terry model and display the model ratings in Elo-scale. More Info | URL | Free |
| Artificial Analysis | 🌟 Independent AI benchmarking for choosing models and API providers. Compare capability, cost per task, output speed and latency, with dedicated evaluations for coding agents, images, video and speech. Published methods and capability–cost charts help you shortlist models before testing your own tasks. Public leaderboards and a free data API are available, alongside commercial data services. More Info | Website | Free/Paid |
| LiveCodeBench | LiveCodeBench is a holistic and contamination-free evaluation benchmark of LLMs for code, built by UC Berkeley + MIT + Cornell, that continuously collects new problems over time. Particularly, LiveCodeBench also focuses on broader code-related capabilities, such as self-repair, code execution, and test output prediction, beyond mere code generation. | URL | Free |
| StructEval | StructEval is a TMLR 2025 benchmark and public leaderboard by University of Waterloo-led TIGER-AI-Lab for evaluating LLM structured-output generation and conversion across 2,035 examples, 18 text and renderable formats, and format-specific structural and visual checks. | Leaderboard Paper GitHub | Free |
| LLM Stats | LLM Stats, an independent, comprehensive LLM leaderboard founded by Jonathan Chavez & Sebastian Crossa (Y Combinator-backed), benchmarks and compares API models using daily‑updated, open‑source community data on capability, price, speed, and context length. | URL | Free |
| ModelCap | ModelCap, an independent platform, publishes live AI model rankings (the ModelCap Index) compiled only from named public sources — benchmark boards such as SWE-bench, ARC-AGI and BFCL, Arena ratings and provider prices — with an evidence tier and score interval per model, per-benchmark leaderboards, head-to-head model comparisons, provider pricing pages and a CC BY 4.0 JSON/CSV dataset. | URL | Free |
| Price Per Token | Compare LLM API pricing across 200+ models from OpenAI, Anthropic, Google, and more, built by Alex Ellman. Includes token counters, cost calculators, and benchmark comparisons. | URL | Free |
| BenchGecko | The data layer of the AI economy — an independent AI model benchmark leaderboard with cross-provider pricing comparison across hundreds of providers. Tracks thousands of models across 128 benchmarks, AI economy indicators, agent leaderboard, and MCP server directory. Free API. | URL | Free |
| llmrun | llmrun is a free site that shows which open LLMs fit your GPU, Mac or mini PC: VRAM needed per quantization and estimated speed (tok/s), plus a composite score fitted with 2PL IRT over public benchmark results (Epoch AI, LiveBench, SWE-bench, MMLU-Pro), with the methodology published. | URL | Free |
| The Aggregate | A free, independent LLM leaderboard that combines public benchmark leaderboards into one daily ranking: a robust IRT model puts every model on an Elo scale with standard errors, every score links to the leaderboard it came from, and any two models can be compared head to head. Also offers a free, no-key MCP server for assistants. | URL | Free |
| Jev Decision Index | A Hugging Face Space tracking open-source reproductions and alternatives to TypeSafe's Jev (System One / RLCD). Combines a Decision Index leaderboard scoring models on structured-decision benchmarks with a News tab cataloging decoding tricks, diffusion-based approaches, and trained decision heads. | Hugging Face | Free |
AI Agent
| Name | Description | Links | Fees |
|---|---|---|---|
| Claude Code | 🌟 Anthropic's AI agent for code, file, and terminal tasks. Available in the terminal CLI, VS Code/Cursor/JetBrains IDE plugins, desktop app, and web. Built on the Claude model family, it offers long-context codebase understanding, multi-file editing, shell command execution, MCP extensions, skills/hooks/subagents, CLAUDE.md project rules, and automatic memory. Supports Remote Control to continue local sessions from your phone or browser. More Info Comparison | GitHub | Free/Paid |
| Codex | 🌟 OpenAI's open-source (Apache-2.0) AI agent with a terminal interface, written in Rust. Supports the terminal CLI, VS Code/Cursor/Windsurf IDE extensions, Codex App/desktop app, and Codex Web cloud tasks. It can understand and modify code directly, with kernel-level sandboxing, MCP, skills/plugins/hooks, subagents, and Remote SSH. Active sessions sync across phone/desktop/web via the ChatGPT relay. More Info Comparison | GitHub  | Free/Paid |
| Kimi Code | 🌟 Moonshot AI's AI agent CLI that runs in your terminal — stands out with native video input and first-class Chinese-language support. Reads and edits code, runs shell commands, searches files, fetches web pages, and plans next steps based on feedback. Supports AI-native MCP configuration, subagents, lifecycle hooks, and ACP editor integration (Zed, JetBrains). More Info Comparison | GitHub  | Free |
| OpenCode | Open-source terminal-native AI agent. Provider-agnostic (Anthropic, OpenAI, Google, local models), with a TUI client/server architecture, LSP integration, and support for custom agents and MCP servers. More Info | URL GitHub  | Free |
| oh-my-pi | Fork of Pi by @can1357. A terminal-native AI agent with the IDE wired in: LSP, DAP debugging, Python/Bun code execution, 40+ providers, and 32 built-in tools. More Info | GitHub  | Free |
| mu | Open-source (MIT) AI agent built on Pi, with a CLI and a desktop app for macOS, Windows and Linux (the app is based on AionUi, Apache-2.0). It adds a judgment kernel: at more than 30 decision points a small, fast judge model (Jev by default; a local 322M judge, a classifier or any LLM can take a point instead) makes the routine calls, such as which chunks of a long tool output enter the context, whether a rule-flagged command was asked for, and whether a fetched page or MCP result carries instructions aimed at the model, so the main model keeps its attention for the work. More Info | GitHub  | Free |
| Manus | 🌱 An action engine that goes beyond answering questions to executing tasks and automating workflows. Rebuilt Manus 2.0 (Cascade framework, always-on cloud computer) plus the Cue personal agent shipped Sep 28, 2026. More Info | URL | Free Trial/Paid |
| DeepSeek Harness | 🌱 DeepSeek's open-source (MIT) agent harness — a full agent runtime (CLI + Web UI) built on an "everything is a plugin" architecture with 40+ built-in packages (LLM, MCP, sandbox, skills, subagents, web, shell, etc.). Currently developer preview — breaking changes expected. More Info | GitHub  | Free |
| Devin | Cognition's autonomous AI software engineer — the pioneer of the assign-a-ticket model: hand tasks to Devin via Slack, Jira or Linear, and it plans, writes code in its own cloud sandbox, runs tests, and opens PRs asynchronously in parallel sessions. Free tier; Core from $20/mo metered by ACUs; Team $500/mo. Windsurf merged into Devin as Devin Desktop (June 2026). More Info | URL | Free/Paid |
| Pi | Open-source (MIT) self-extensible AI agent by earendil-works. Modular monorepo with a polished interactive CLI, a unified multi-provider LLM layer, and a plugin architecture; emphasizes supply-chain hardening and sandboxed execution (Gondolin, Docker, OpenShell). More Info | GitHub  | Free |
| Grok build | xAI's terminal-based AI agent, open-sourced and written in Rust. Runs as a full-screen, mouse-interactive TUI that understands your codebase, edits files, runs shell commands, searches the web, and manages long-running tasks. Supports interactive, headless (scripting/CI), and ACP editor-embedded modes, with MCP servers, skills, plugins, hooks, and sandboxing. More Info | GitHub  | Free |
| OpenClaw | Open-source, self-hosted AI agent that runs locally and autonomously executes tasks. Connects to WhatsApp, Telegram, Slack, Discord and more, with browser control, system access, and persistent memory. Now maintained by the independent OpenClaw Foundation with OpenAI support; one of the fastest-growing open-source projects. More Info | GitHub  | Free |
| OpenWorker | Open-source, local-first desktop AI coworker by Andrew Ng. Delivers finished work (documents, Slack replies, calendar updates, reports) rather than chat answers. Supports 25+ app connectors, MCP, and bring-your-own-key models (OpenAI, Anthropic, Gemini, DeepSeek, Kimi, Qwen, Ollama, etc.) with local storage of data and credentials. More Info | GitHub  | Free |
| autoresearch | An autonomous AI agent for machine learning research by Karpathy. It can formulate hypotheses, modify training code, run experiments, and evaluate results autonomously to find performance improvements. More Info | GitHub  | Free |
| ml-intern | An open-source, autonomous AI agent by Hugging Face designed to function as a specialized Machine Learning Engineer. It handles the end-to-end ML lifecycle, including researching papers, writing code, running experiments, and shipping models to the Hugging Face Hub. More Info | GitHub  | Free |
| msitarzewski/agency-agents | Open-source AI agency toolkit with 230+ specialized agent personas across engineering, design, marketing, sales, finance, game dev, healthcare, and more. Features ready-to-use agent personalities, workflow templates, deliverables, and a native desktop app to browse/install agents into Claude Code, Cursor, Codex, Gemini, Kimi Code, Aider, and other AI dev tools. More Info | GitHub  | Free |
| Hermes Agent | Open-source agentic framework built by NousResearch, fine-tuned on Hermes models with strong tool calling and reasoning capabilities for building autonomous AI agents. More Info | GitHub  | Free |
| DeepTutor | An agent-native learning workspace by HKUDS that unifies AI tutoring, problem solving, quiz generation, research, visualization, and mastery practice. Features multi-engine RAG, persistent memory, Partners/My Agents, and live coding CLI integration. More Info | GitHub  | Free |
| potpie-ai/potpie | Open Source AI Agents for your codebase in minutes. Use pre-built agents for Q&A, Testing, Debugging and System Design or create your own purpose-built agents. More Info | URL , GitHub  | Free Trial |
| CLI-Anything | An open-source framework that makes all software agent-native — auto-generates production-ready CLIs for any application via a 7-phase pipeline, enabling AI agents to control any desktop software. Supports 16+ AI agent platforms including Claude Code, OpenClaw, Codex and includes CLI-Hub for community CLI discovery. More Info | GitHub  | Free |
Coding Agents
Coding editors, IDEs, and app builders with AI assistance and agentic development workflows.
| Name | Description | Links | Fees |
|---|---|---|---|
| Cursor | 🌟 The leading AI coding editor: switch between OpenAI, Anthropic, Gemini, Grok and Cursor's own Composer models, and run parallel agents across desktop, CLI, Slack and GitHub. A wholly owned SpaceX (SpaceXAI) subsidiary since Aug 2026. More Info | URL | Paid/Free Trial |
| Trae | ByteDance's AI coding IDE built around agentic programming: AI Q&A, code auto-completion, and autonomous multi-step coding tasks. More Info | URL | Free |
| GitHub Copilot | GitHub's code writing assistant, now a multi-vendor model picker (OpenAI, Anthropic, Gemini and more) chosen per task More Info | URL | Free/Paid |
| Antigravity | Google's agent-first AI coding platform, available as a desktop app, IDE, terminal CLI (agy), and SDK. Supports Gemini 4 Argon and Gemini 3.8 Flash, alongside third-party models such as Claude and GPT-OSS; model access depends on the plan. Editor/Manager views coordinate multiple agents with a browser, terminal, verifiable Artifacts, subagents, hooks, and scheduled tasks. More Info | URL | Free/Paid |
| Bolt.new | StackBlitz's browser-based AI full-stack app builder — describe an app in plain language and the AI agent scaffolds, runs, and deploys it in the browser via WebContainers (real Node.js at near-native speed, no local setup or cloud VM). Free / Pro $25/mo / Teams $30 per member / Enterprise plans; open-source sibling bolt.diy supports bring-your-own-key. One of the fastest-growing dev tools ($40M ARR within 5 months of launch; Microsoft Azure partnership since May 2026). More Info | URL | Free/Paid |
Code Review & Development Utilities
Code review, debugging, development conventions, configuration, and companion tools.
| Name | Description | Links | Fees |
|---|---|---|---|
| git-lrc | Free, Unlimited AI Code Reviews That Run on Every Commit | GitHub  | Free |
| alibaba/open-code-review | Open-source AI code-review CLI/agent by Alibaba. Reviews Git diffs with deterministic pipelines plus an LLM agent, supports line-level comments, configurable OpenAI/Anthropic-compatible endpoints, custom rules, JSON CI output, WebUI session viewer, and Claude Code/Codex integrations. | GitHub  | Free |
| Kodus | Open Source Code Review Agent | GitHub  | Free/Paid |
| AI-Codereview-Gitlab | GitLab automatic code review tool based on LLMs (DeepSeek, OpenAI, etc.). Supports DingTalk/WeCom/Feishu notifications and daily reports generation. Docker deployment supported with visual Dashboard. | GitHub  | Free |
| code-review-graph | Builds a structural code graph with Tree-sitter, tracks incremental changes, and exposes precise review context to AI coding assistants via MCP to reduce redundant codebase reads. | GitHub  | Free |
| dbForge AI Assistant | AI-powered tool that generates, optimizes, and troubleshoots SQL code; indispensable for developers, DBAs, and analysts | URL | Paid/Free Trial |
| scalene | Scalene: a high-performance, high-precision CPU, GPU, and memory profiler for Python with AI-powered optimization proposals | GitHub  | Free |
| OpenChamber | Desktop and web GUI for the OpenCode AI agent. Provides session management, diffs, and workspace control on top of OpenCode. | URL GitHub  | Free |
| Happy Coder | Mobile and Web client for Codex and Claude Code, with realtime voice, encryption and fully featured | URL GitHub  | Free |
| Termux | Android terminal emulator and Linux environment app that allows you to run coding tools, AI models and development environments directly on your mobile device, with built-in SSH client support for logging into remote hosts. | GitHub  | Free |
| Cate | Open source desktop IDE on an infinite zoomable canvas. You arrange Monaco editors, xterm.js terminals, embedded browsers, and Claude Code agent panels in freeform zoomable space instead of tabs. Panels can float, dock, or detach into separate windows, and layout persists per project. Built with Electron, runs on macOS, Windows, and Linux. | URL GitHub  | Free |
| CLIProxyAPI | A proxy server that exposes AI CLI tools and subscriptions (Claude Code, Codex, Gemini, Grok, etc.) through OpenAI/Gemini/Claude/Codex-compatible API interfaces. Supports OAuth login, streaming/WebSocket responses, function calling, multimodal input, and multi-account load balancing across providers. | GitHub  | Free |
| Markstream | Open-source streaming Markdown renderer family for AI chat interfaces. Handles incomplete Markdown and smooth token pacing, with maintained Vue 3/Nuxt, React/Next.js, Svelte, Angular, Vue 2, and framework-agnostic parser/core packages. Supports Mermaid, KaTeX, code highlighting, safe HTML, and SSR. | Frameworks Vue Playground GitHub  | Free |
| Cortex | Open-source CLI that generates interactive API documentation, typed SDKs, and MCP servers from OpenAPI, AsyncAPI, GraphQL, gRPC, OpenRPC, and Markdown. | GitHub  | Free |
| ECC | Open-source enhancement framework for AI coding agents. Unified agent/skills/rules/Hooks/MCP configuration across Claude Code, Codex, Cursor, OpenCode, Gemini, Zed, GitHub Copilot. Ships with 60 sub-agents, 232 skills, persistent memory, AgentShield security scan, and desktop dashboard GUI. | GitHub  | Free |
| CC Switch | Cross-platform desktop app to manage Claude Code, Codex, Gemini CLI, OpenCode, and OpenClaw. Features visual configuration for API providers, MCP servers, and skills with bidirectional sync. | GitHub  | Free |
| codex-plugin-cc | Claude Code plugin that enables using OpenAI Codex for code reviews and task delegation directly within Claude Code workflow. Provides /codex:review, /codex:adversarial-review and other commands | GitHub  | Free |
| AGENTS.md | An open standard for codebase documentation designed specifically for AI coding agents. Provides structured context, instructions, and constraints to help agents work more effectively and safely. | GitHub  | Free |
| Spec Kit | GitHub's open-source toolkit for Spec-Driven Development (SDD). Provides the specify CLI and slash commands (/speckit.specify, /speckit.plan, /speckit.implement, etc.) that integrate with 30+ AI coding agents to turn specifications into executable implementation plans. | GitHub  | Free |
| musistudio/claude-code-router | A powerful tool to route Claude Code requests to different models and customize any request. Supports multi-provider routing, request/response transformation, dynamic model switching, CLI model management, and GitHub Actions integration. | GitHub  | Free |
Agent Workspaces & Collaboration
Manage and schedule existing agents, tasks, sessions, workspaces, and team collaboration.
| Name | Description | Links | Fees |
|---|---|---|---|
| OpenDots | 🌱 CopilotKit's open-source self-hosted agent workspace template (open alternative to OpenAI Dots). Gives each Dot its own computer with browser/file/terminal permissions, MCP tools, Slack/calls, Spaces document collaboration, background tasks, and human-in-the-loop approval. Requires self-hosting and connecting CopilotKit Intelligence + an OpenAI-compatible model. More Info | GitHub  | Free |
| Multica | Open-source managed agents platform to integrate AI coding agents (Claude Code, Codex, Cursor, Copilot, OpenCode, OpenClaw, Hermes, Gemini, Pi, Kimi, Kiro CLI, and more) into dev teams as first-class teammates. Features task assignment, progress tracking, reusable skills, and local daemon auto-detection of installed agent CLIs. | GitHub  | Free/Paid |
| Paperclip | Open-source AI agent orchestration platform that models AI operations as a company. Define org charts, reporting lines, budgets, and governance policies, then let agents work autonomously within that structure. Supports OpenClaw, Claude Code, Codex, and other agent backends. Heartbeat-based execution with cost controls and audit trails. | GitHub  | Free |
| Symphony | OpenAI's open-source project-to-delivery harness. Turns project requirements into isolated, autonomous implementation runs; agents monitor boards like Linear, open PRs, and merge code with proof-of-work (CI status, review feedback, demo videos). Currently in engineering preview. | GitHub  | Free |
| QM | YC open-source multi-agent harness for organizations. Runs on Slack and Web with scoped workspaces (memory, files, credentials, permissions) per employee/project/channel; headless core with replaceable models/harnesses and self-hostable via Docker/Fly/AWS. | GitHub  | Free |
| Buzz | Open-source, self-hostable workspace where humans and AI agents share the same rooms. Agents get their own keys and audit trail to open repos, send patches, review code, run workflows, edit canvases, and join voice huddles. Built on a Nostr relay so every message and action is a signed event. | GitHub  | Free |
| Orkas | Open-source, local-first multi-agent AI desktop client. A Commander decomposes goals and coordinates specialist agents in parallel or sequence, with private skills and memory. Supports BYO model providers and local model endpoints, and can drive Claude Code, Codex, OpenClaw, OpenCode, and Hermes. Cross-platform (macOS / Windows / Linux). | Website GitHub  | Free |
| Orca | Cross-platform Agent Development Environment (ADE) for orchestrating fleets of parallel AI coding agents. Run Claude Code, Codex, OpenCode, Pi, and any CLI agent side-by-side in isolated git worktrees, compare outputs, and merge the winner. Features mobile companion, terminal splits, Design Mode for capturing Chromium UI elements, GitHub/Linear integration, SSH worktrees, AI diff annotations, and Computer Use. | Website GitHub  | Free |
| Paseo | 🌱 Open-source agentic development environment that unifies Claude Code, Codex, Copilot, OpenCode, Pi, Antigravity, and Muse Code in one interface. Runs parallel agents in isolated worktrees with built-in orchestration, editor, terminals, diffs, PRs, browser, and a plugin system. Self-hosted, privacy-first, cross-platform desktop/mobile/web/CLI, plus a TypeScript SDK. More Info | GitHub | Free |
| AgentsMesh | The AI Agent Workforce Platform. Self-hostable remote AI workstations (AgentPods) with PTY sandbox + git worktree isolation, multi-agent collaboration via channels and pod bindings, built-in Kanban with MR/PR integration, per-pod MCP server. Supports Claude Code, Codex CLI, Gemini CLI, Aider, OpenCode. | GitHub  | Free (BSL-1.1) |
| Kortix | The open-source AI Management System — run a company's agent workforce with agents, skills, memory and 3,000+ connectors as files in one git repo you own; each session boots its own isolated Linux machine, and work lands through a human-reviewed change request. Self-host or managed cloud, any model with your own keys. | GitHub  | Free |
Agent Frameworks & Runtimes
Build agents, define execution workflows, and provide agent runtimes.
| Name | Description | Links | Fees |
|---|---|---|---|
| Microsoft Agent Framework | Microsoft's MIT-licensed framework for production AI agents and multi-agent workflows, and the recommended successor to archived AutoGen. Supports Python and .NET, multiple model providers, graph-based orchestration, checkpointing, human approval, and OpenTelemetry observability. Suitable for teams that need durable, inspectable workflows from local development to cloud deployment; model and hosting costs depend on the chosen services. | GitHub  | Free |
| LangGraph | LangChain's low-level orchestration framework for building stateful, multi-agent LLM applications as graphs. Models workflows as nodes and edges with persistent state, checkpointing, and human-in-the-loop interrupt/resume, making long-running, fault-tolerant agents easier to build, debug, and observe. Ships graph/state primitives, a Studio visual debugger, and LangGraph Platform for durable execution, persistence, and streaming at scale. Works with any model provider; Python and JS/TS. More Info | GitHub  Website | Free |
| MastraAI | Mastra is an opinionated TypeScript framework that helps you build AI applications and features quickly. It gives you the set of primitives you need: workflows, agents, RAG, integrations and evals | GitHub  | Free |
| agentscope | Agent-Oriented Programming for Building LLM Applications, Open-sourced by Alibaba | GitHub | Free |
| FlowGram.AI | An extensible visual AI workflow development framework with built-in canvas, form engine, variable scope chain, and ready-to-use materials (LLM, Condition, Code Editor, etc.) for building AI workflow platforms faster. | GitHub  | Free |
| Conductor | Netflix-originated open-source durable workflow engine now focused on AI agent orchestration. Provides event-driven execution, durable state, retries/timeouts, 14+ LLM providers, MCP tool calling, function calling, human-in-the-loop approval, and vector DB integration for RAG. Trusted in production at Netflix, Tesla, LinkedIn, and J.P. Morgan. | GitHub  | Free |
| deer-flow | ByteDance's open-source long-horizon SuperAgent harness. Orchestrates sub-agents, memory, sandboxes, tools, skills, and message gateways to research, code, and create. Built on LangGraph/LangChain with a web UI and multi-provider LLM support. | GitHub  | Free |
| AX | Google's open-source agentic orchestration runtime — positioned as "the Kubernetes for agents". Declarative YAML manifests (Workspace, Task, Model) with sandboxed execution, CPU/memory limits, ax suspend/resume and SSH debugging; single-cluster design targets billions of autonomous agent workloads. Early stage, breaking changes expected.More Info | GitHub  | Free |
Browser & Computer Automation
Automate web, desktop, and mobile interactions with supporting browsers and sandboxes.
| Name | Description | Links | Fees |
|---|---|---|---|
| browser-use | Open-source (MIT) Python library that lets AI agents control web browsers: natural-language tasks, DOM extraction, form filling and multi-step navigation, with any LLM. Browser Use Cloud adds hosted stealth Chromium you can drive over CDP (also from Playwright or Puppeteer), residential proxies, CAPTCHA solving, persistent profiles, a browser-agent API and a remote MCP server, billed per browser-hour with no subscription. More Info | GitHub  Website | Free/Paid |
| Cua | Open-source infrastructure for Computer-Use Agents (CUA). Provides sandboxes, drivers, and SDKs for macOS, Windows, Linux, and Android, enabling AI to control full desktops just like a human. | GitHub  | Free |
| ARTEMIS | Google's open-source framework that turns natural-language instructions into reliable Android automation: cross-app task execution, multimodal targeting, and Flash (fast reactive) / Pro (multi-agent planning) execution modes. Ships a Web UI console, MCP server, Python SDK, and CLI for integration with Claude Code, Codex and other agents. More Info | GitHub  | Free |
| Obscura | Open-source (Apache-2.0) headless browser engine written in Rust for AI agents and web scraping. A drop-in replacement for headless Chrome via the Chrome DevTools Protocol, compatible with Puppeteer/Playwright: ~30 MB memory, instant startup, and a built-in stealth mode with anti-fingerprinting and tracker blocking. | GitHub  | Free |
| Steel Browser | Open-source browser sandbox and automation infrastructure for AI agents and apps with session-backed workflows, screenshots, PDFs, proxies, and anti-bot tooling. | GitHub  | Free |
| Lightpanda | A lightweight, high-performance headless browser written in Zig, designed for AI agents. Optimized for speed and low memory (~9x faster than Chrome), featuring native MCP support and Markdown output for LLMs. | GitHub  | Free |
| bb-browser | Enable AI agents to control your real Chrome browser session, accessing sites with your existing login state without APIs. | GitHub  | Free |
| Chrome DevTools MCP | Official Model Context Protocol server from Chrome DevTools that enables AI coding agents (like Claude Code, Gemini CLI) to control and inspect a live Chrome browser. Supports performance trace analysis, network request debugging, screenshot capture, and console log inspection, helping AI agents better run, test, and debug code. | GitHub  | Free |
Agent Memory & Context
Store and retrieve historical memory, and build context from codebases and knowledge sources.
| Name | Description | Links | Fees |
|---|---|---|---|
| MemPalace | Open-source AI memory system that stores full conversations and project data locally without cloud dependencies. Organizes memories into a hierarchical "palace" structure, achieves 96.6% recall on LongMemEval — the highest published score for a fully local setup that calls no APIs, supports MCP integration and works offline with local LLMs. | GitHub  | Free |
| TencentDB Agent Memory | Tencent Cloud's open-source memory infrastructure for agent teams. Provides Memory Hub for cross-session/cross-agent memory sharing, Chat Memory with L0–L3 distillation, Skill library for reusable expertise, and Wiki/CodeGraph for docs/code knowledge maps. Works with Claude Code, CodeBuddy, and other agents. | GitHub  | Free |
| Screenpipe | Searchable screen-text and audio history for recall, meeting notes, and work summaries via MCP or a local API. Raw history stays local by default; configured cloud AI, transcription, sync, and integrations can process context off-device. Source-available under the Screenpipe Commercial License. macOS, Windows, and Linux. | GitHub  | Limited free plan; paid subscriptions |
| codesight | CLI token optimizer and AI context generator with built-in MCP server. Scans codebases to extract routes, schema, components, and dependencies, reducing AI context tokens by 9x–13x. Profiles for Claude Code, Cursor, Copilot, Codex, and Windsurf. Zero runtime dependencies. | GitHub  | Free |
| Understand Anything | AI-powered codebase and knowledge base analysis tool that builds interactive visual knowledge graphs. Features multi-agent analysis, architecture guided tours, and semantic search for large codebases. Works as a plugin for Claude Code, Cursor, and more. | GitHub  | Free |
Agent Skills
| Name | Description | Links | Fees |
|---|