# AgentSpot — full listing index

> A curated, automatically refreshed directory where you can discover which AI agents, reusable skills, and agent tools are popular — ranked by transparent public metrics.
> Generated 2026-09-23. Data license: CC-BY-4.0 (https://creativecommons.org/licenses/by/4.0/). Attribution: AgentSpot (https://agentspot.dev).

## Agents

- **OpenClaw** (Other (see repo)) — Open-source general-purpose personal agent that runs on your own machine and lives in your messaging apps — with tools, persistent memory, and automations wherever you already talk. [390,313 stars, 2,489,212 npm downloads/7d] | Can run locally | Install: `npm install openclaw` Page: https://agentspot.dev/agents/openclaw/ | Repo: https://github.com/openclaw/openclaw | Site: https://openclaw.ai
- **hermes-agent** (MIT) — Nous Research's open-source personal agent — persistent memory, tool use, and multi-channel deployment — among the most heavily used open agents by independent inference-volume measures. [248,251 stars, 33,743 pypi downloads/7d] | Can run locally | Install: `pip install hermes-agent` Page: https://agentspot.dev/agents/hermes-agent/ | Repo: https://github.com/NousResearch/hermes-agent | Site: https://hermes-agent.nousresearch.com
- **langchain** (MIT) — A widely adopted framework for building LLM applications and agents, with composable abstractions for prompts, memory, tool calling, and retrieval across many model providers. [146,917 stars, 39,631,410 pypi downloads/7d] | Can run locally | Install: `pip install langchain` Page: https://agentspot.dev/agents/langchain/ | Repo: https://github.com/langchain-ai/langchain | Site: https://docs.langchain.com/langchain/
- **browser-use** (MIT) — Lets AI agents control a real web browser to navigate sites and complete tasks, turning natural-language goals into automated clicks, form fills, and data extraction. [116,034 stars, 1,756,111 pypi downloads/7d] | Can run locally | Install: `pip install browser-use` Page: https://agentspot.dev/agents/browser-use/ | Repo: https://github.com/browser-use/browser-use | Site: https://browser-use.com
- **TradingAgents** (Apache-2.0) — A multi-agent framework for LLM-driven financial trading research: specialist analyst agents for fundamentals, sentiment, news and technicals feed a bull-versus-bear debate, then risk management and a trader agent synthesise a decision. Model providers and market-data vendors are pluggable. [108,247 stars, 376 pypi downloads/7d] | Can run locally | Install: `pip install tradingagents` Page: https://agentspot.dev/agents/tradingagents/ | Repo: https://github.com/TauricResearch/TradingAgents | Site: https://arxiv.org/pdf/2412.20138
- **langgraph** (MIT) — A library from the LangChain team for building stateful, multi-step agent workflows as graphs, with durable execution, branching, and human-in-the-loop control. [42,171 stars, 9,995,800 pypi downloads/7d] | Can run locally | Install: `pip install langgraph` Page: https://agentspot.dev/agents/langgraph/ | Repo: https://github.com/langchain-ai/langgraph | Site: https://docs.langchain.com/oss/python/langgraph/
- **crewAI** (MIT) — A framework for orchestrating teams of role-playing autonomous agents that collaborate on complex tasks, each with its own goal, tools, and backstory. [58,935 stars, 540,946 pypi downloads/7d] | Can run locally | Install: `pip install crewai` Page: https://agentspot.dev/agents/crewai/ | Repo: https://github.com/crewAIInc/crewAI | Site: https://crewai.com
- **AutoGPT** (Other (see repo)) — One of the earliest autonomous GPT agents, now a platform for building, deploying, and running AI agents that chain tasks toward a goal with minimal human input. [187,505 stars, 93 docker downloads/7d] | Can run locally | Install: `docker pull significantgravitas/auto-gpt` Page: https://agentspot.dev/agents/autogpt/ | Repo: https://github.com/Significant-Gravitas/AutoGPT | Site: https://agpt.co
- **autogen** (CC-BY-4.0) — Microsoft's framework for building multi-agent AI applications, where several conversable agents coordinate, call tools, and hand off work to solve tasks together. [61,116 stars, 93,671 pypi downloads/7d] | Can run locally | Install: `pip install autogen-agentchat` Page: https://agentspot.dev/agents/autogen/ | Repo: https://github.com/microsoft/autogen | Site: https://microsoft.github.io/autogen/
- **agno** (Apache-2.0) — A Python framework and platform for building, running, and managing AI agents, with built-in memory, tools, and multi-agent teams. [42,316 stars, 382,043 pypi downloads/7d] | Can run locally | Install: `pip install agno` Page: https://agentspot.dev/agents/agno/ | Repo: https://github.com/agno-agi/agno | Site: https://docs.agno.com
- **OpenAI Agents SDK** (MIT) — OpenAI's lightweight Python framework for multi-agent workflows, with handoffs between agents, guardrails, and built-in tracing. [29,652 stars, 2,915,690 pypi downloads/7d] | Can run locally | Install: `pip install openai-agents` Page: https://agentspot.dev/agents/openai-agents/ | Repo: https://github.com/openai/openai-agents-python | Site: https://openai.github.io/openai-agents-python/
- **nanobot** (MIT) — Ultra-lightweight self-hosted personal agent framework in Python from HKUDS, with a web UI, tools, memory, MCP support, and multi-agent workflows. [48,503 stars, 9,458 pypi downloads/7d] | Can run locally | Install: `pip install nanobot-ai` Page: https://agentspot.dev/agents/nanobot/ | Repo: https://github.com/HKUDS/nanobot | Site: https://nanobot.wiki
- **cua** (MIT) — Open-source infrastructure for computer-use agents: sandboxes, SDKs, and benchmarks for AI that controls full macOS, Linux, and Windows desktops. [26,053 stars, 770 docker downloads/7d] | Can run locally | Install: `docker pull trycua/cua-xfce` Page: https://agentspot.dev/agents/cua/ | Repo: https://github.com/trycua/cua | Site: https://cua.ai
- **pydantic-ai** (MIT) — An agent framework from the Pydantic team that brings type-safe, structured outputs and validation to LLM applications and tool calling. [20,128 stars, 1,201,992 pypi downloads/7d] | Can run locally | Install: `pip install pydantic-ai` Page: https://agentspot.dev/agents/pydantic-ai/ | Repo: https://github.com/pydantic/pydantic-ai | Site: https://pydantic.dev/pydantic-ai
- **deer-flow** (MIT) — An open-source long-horizon agent harness from ByteDance that researches, writes code, and produces content using sandboxes, memory, tools, and subagents. [82,888 stars] | Can run locally Page: https://agentspot.dev/agents/deer-flow/ | Repo: https://github.com/bytedance/deer-flow | Site: https://deerflow.tech
- **Vibe-Trading** (MIT) — Personal trading agent framework from HKUDS: LLM-driven market analysis, strategy backtesting, and automated execution under one agent loop. [33,870 stars, 1,110 pypi downloads/7d] | Install: `pip install vibe-trading-ai` Page: https://agentspot.dev/agents/vibe-trading/ | Repo: https://github.com/HKUDS/Vibe-Trading | Site: https://vibetrading.wiki/
- **anything-llm** (MIT) — An all-in-one, self-hostable desktop and web app for running local-first AI agents over your own documents, with RAG, tools, and multi-user support. [66,361 stars, 85,178 docker downloads/7d] | Can run locally | Install: `docker pull mintplexlabs/anythingllm` Page: https://agentspot.dev/agents/anything-llm/ | Repo: https://github.com/Mintplex-Labs/anything-llm | Site: https://anythingllm.com
- **letta** (Apache-2.0) — A platform for building stateful agents with long-term memory that persists across sessions, letting agents learn and improve over time (formerly MemGPT). [24,856 stars, 3,821 pypi downloads/7d] | Can run locally | Install: `pip install letta` Page: https://agentspot.dev/agents/letta/ | Repo: https://github.com/letta-ai/letta | Site: https://docs.letta.com/
- **goose** (Apache-2.0) — Open-source agent that runs on your own machine as a desktop app or terminal CLI and does real work rather than suggesting it — installing dependencies, editing and testing code, and driving external systems through MCP extensions, with any model you point it at. Built at Block and now governed by the Linux Foundation's Agentic AI Foundation. [54,584 stars] | Can run locally Page: https://agentspot.dev/agents/goose/ | Repo: https://github.com/aaif-goose/goose | Site: https://goose-docs.ai/
- **cherry-studio** (AGPL-3.0) — Cross-platform AI studio bundling chat, autonomous agents, and hundreds of ready-made assistants over unified access to frontier and local models. [52,107 stars] | Can run locally Page: https://agentspot.dev/agents/cherry-studio/ | Repo: https://github.com/CherryHQ/cherry-studio | Site: https://cherryai.com
- **CowAgent** (MIT) — Agent harness grown out of the chatgpt-on-wechat project: plans tasks, runs tools and skills, and carries persistent memory across many models and chat platforms. [47,092 stars, 4,195 docker downloads/7d] | Can run locally | Install: `docker pull zhayujie/cowagent` Page: https://agentspot.dev/agents/cowagent/ | Repo: https://github.com/zhayujie/CowAgent | Site: https://cowagent.ai
- **CopilotKit** (MIT) — Frontend framework for building agent-native applications in React, Angular, and beyond — generative UI, in-app copilots, and the AG-UI protocol connecting interfaces to any agent backend. [37,496 stars] | Can run locally Page: https://agentspot.dev/agents/copilotkit/ | Repo: https://github.com/CopilotKit/CopilotKit | Site: https://docs.copilotkit.ai
- **Mastra** (Other (see repo)) — A TypeScript framework for building AI applications and agents, with workflows, evals, memory, and MCP support in one toolkit. [28,283 stars] | Can run locally Page: https://agentspot.dev/agents/mastra/ | Repo: https://github.com/mastra-ai/mastra | Site: https://mastra.ai
- **activepieces** (Other (see repo)) — Open-source workflow automation platform where AI agents, MCP servers, and hundreds of prebuilt integration pieces compose into automated business flows that agents can drive directly. [24,682 stars, 21,187 docker downloads/7d] | Can run locally | Install: `docker pull activepieces/activepieces` Page: https://agentspot.dev/agents/activepieces/ | Repo: https://github.com/activepieces/activepieces | Site: https://www.activepieces.com
- **Vercel AI SDK** (Other (see repo)) — Vercel's TypeScript toolkit for building AI applications: a unified interface across model providers, plus streaming, tool calling, and generative UI for React, Svelte, and Vue. [26,911 stars] | Can run locally Page: https://agentspot.dev/agents/vercel-ai/ | Repo: https://github.com/vercel/ai | Site: https://ai-sdk.dev
- **SuperAGI** (MIT) — A developer-first open-source framework for building, running, and managing autonomous AI agents, with a GUI, tool integrations, and concurrent agent support. [17,687 stars] | Can run locally Page: https://agentspot.dev/agents/superagi/ | Repo: https://github.com/TransformerOptimus/SuperAGI | Site: https://superagi.com/
- **Cloudflare Agents** (MIT) — Cloudflare's SDK for building and deploying agents on its edge platform, using Durable Objects for state and workflows that survive restarts. [5,647 stars] | Needs a hosted service Page: https://agentspot.dev/agents/cloudflare-agents/ | Repo: https://github.com/cloudflare/agents | Site: https://developers.cloudflare.com/agents/

## Skills

- **ponytail** (MIT) — Agent skill that makes coding agents write like a pragmatic senior engineer — minimal diffs, no speculative abstractions — via a drop-in ruleset supported across Claude Code, Codex, and a dozen-plus other harnesses. [144,725 stars, 15,064 npm downloads/7d] | Install: `npm install @dietrichgebert/ponytail` Page: https://agentspot.dev/skills/ponytail/ | Repo: https://github.com/DietrichGebert/ponytail | Site: https://ponytail.dev
- **ECC (Everything Claude Code)** (MIT) — A comprehensive configuration pack for coding agents — Claude Code, Codex, Cursor, and others — bundling skills, hooks, commands, and memory into a plan-test-implement-review-verify workflow. [265,799 stars] Page: https://agentspot.dev/skills/ecc/ | Repo: https://github.com/affaan-m/ECC | Site: https://ecc.tools
- **skills** (Other (see repo)) — Anthropic's public repository of Agent Skills — packaged instructions and resources (the SKILL.md format) that teach agents to perform specific tasks reliably. [177,749 stars] Page: https://agentspot.dev/skills/skills/ | Repo: https://github.com/anthropics/skills
- **prompts.chat** (Other (see repo)) — A community collection of reusable prompts (formerly Awesome ChatGPT Prompts) that you can share, discover, and self-host for your organization. [171,076 stars] Page: https://agentspot.dev/skills/prompts-chat/ | Repo: https://github.com/f/prompts.chat | Site: https://prompts.chat
- **gstack** (MIT) — Garry Tan's published Claude Code configuration: 23 slash-command skills that role-play a product team — engineering manager, designer, QA, release engineer — chained into a plan-build-review-ship workflow, plus a set of CLI helpers. Also ports to Codex, Cursor, and other coding agents. [133,986 stars] Page: https://agentspot.dev/skills/gstack/ | Repo: https://github.com/garrytan/gstack
- **awesome-claude-skills** (Other (see repo)) — A curated list of Claude Skills, resources, and tools for customizing AI agent workflows. [75,522 stars] Page: https://agentspot.dev/skills/awesome-claude-skills/ | Repo: https://github.com/ComposioHQ/awesome-claude-skills
- **Prompt-Engineering-Guide** (MIT) — A comprehensive open guide to prompt engineering, context engineering, RAG, and agents, with lessons, papers, and worked examples. [78,570 stars] Page: https://agentspot.dev/skills/prompt-engineering-guide/ | Repo: https://github.com/dair-ai/Prompt-Engineering-Guide | Site: https://www.promptingguide.ai/
- **claude-plugins-official** (Apache-2.0) — Anthropic's official, managed directory of high-quality Claude Code plugins, which bundle skills, commands, and MCP servers. [36,645 stars] Page: https://agentspot.dev/skills/claude-plugins-official/ | Repo: https://github.com/anthropics/claude-plugins-official | Site: https://code.claude.com/docs/en/plugins
- **agents** (MIT) — A large community collection of subagents and skills packaged as plugins for Claude Code, Codex CLI, Cursor, and other agent harnesses. [39,899 stars] Page: https://agentspot.dev/skills/agents/ | Repo: https://github.com/wshobson/agents | Site: https://sethhobson.com
- **guizang-ppt-skill** (AGPL-3.0) — Agent skill that turns a prompt or document into polished HTML slide decks — editorial and Swiss layouts, image prompts, and social covers — installable in Claude Code, Codex, and other harnesses. [26,836 stars] Page: https://agentspot.dev/skills/guizang-ppt-skill/ | Repo: https://github.com/op7418/guizang-ppt-skill
- **Cursor Plugins** (Other (see repo)) — Cursor's official plugin collection, and the home of pstack — Lauren Tan's set of engineering skills and principles that routes a request to the right playbook and insists on runtime verification rather than a passing build. The skills are plain SKILL.md files, so they load in Claude Code, Codex and other agents, not only Cursor. [8,450 stars] | Can run locally Page: https://agentspot.dev/skills/cursor-plugins/ | Repo: https://github.com/cursor/plugins | Site: https://cursor.com/docs/plugins
- **Distilly** (MIT) — Agent skill and Python toolkit that distills how a specific person thinks — from source material you supply — into a reusable persona skill any agent can load. Built for capturing a departing colleague's know-how, and stretching to mentors, public figures, and fictional characters. Renamed from Colleague Skill in August 2026. [24,986 stars] Page: https://agentspot.dev/skills/distilly/ | Repo: https://github.com/titanwings/distilly
- **ClawHub** (MIT) — OpenClaw's public registry of skills and plugins: a browsable directory where community authors publish capability packs — file-based planning, mapping and routing, phone calls, Home Assistant control — that an OpenClaw install pulls in on demand. Built on the open Agent Skills format, so entries are portable to other agents that read it. [9,443 stars] Page: https://agentspot.dev/skills/clawhub/ | Repo: https://github.com/openclaw/clawhub | Site: https://clawhub.ai
- **playwright-skill** (MIT) — A model-invoked Claude Code skill for browser automation with Playwright, letting an agent write and run its own end-to-end tests. [3,138 stars] Page: https://agentspot.dev/skills/playwright-skill/ | Repo: https://github.com/lackeyjb/playwright-skill
- **goose-skills** (Other (see repo)) — A library of growth and go-to-market skills for coding agents, covering ads, SEO, content, and lead generation. [1,221 stars] Page: https://agentspot.dev/skills/goose-skills/ | Repo: https://github.com/gooseworks-ai/goose-skills | Site: https://gooseworks.ai/cli

## Tools

- **firecrawl** (AGPL-3.0) — An API that crawls, scrapes, and searches the web at scale and returns clean, LLM-ready markdown for agents and RAG pipelines. [183,612 stars, 1,238,677 pypi downloads/7d] | Install: `pip install firecrawl-py` Page: https://agentspot.dev/tools/firecrawl/ | Repo: https://github.com/firecrawl/firecrawl | Site: https://firecrawl.dev
- **open-webui** (Other (see repo)) — A self-hosted, user-friendly web interface for local and API-based LLMs, with support for Ollama, OpenAI-compatible endpoints, RAG, and tools. [152,892 stars, 315,267 pypi downloads/7d] | Can run locally | Install: `pip install open-webui` Page: https://agentspot.dev/tools/open-webui/ | Repo: https://github.com/open-webui/open-webui | Site: https://openwebui.com
- **Agent-Reach** (MIT) — Command-line tool that gives AI agents eyes across the public internet — reading and searching X/Twitter, Reddit, YouTube, GitHub, Bilibili, and Xiaohongshu from one interface, with no per-platform API keys required. [85,001 stars, 1,149 pypi downloads/7d] | Can run locally | Install: `pip install agent-reach` Page: https://agentspot.dev/tools/agent-reach/ | Repo: https://github.com/Panniantong/Agent-Reach
- **mem0** (Apache-2.0) — A universal memory layer for AI agents and assistants that stores, recalls, and manages long-term user and session memory. [65,878 stars, 448,663 pypi downloads/7d] | Can run locally | Install: `pip install mem0ai` Page: https://agentspot.dev/tools/mem0/ | Repo: https://github.com/mem0ai/mem0 | Site: https://mem0.ai
- **n8n** (Other (see repo)) — A fair-code workflow automation platform with native AI features and hundreds of integrations, letting you wire agents and tools into visual, self-hostable pipelines. [205,751 stars, 2,814,498 docker downloads/7d] | Can run locally | Install: `docker pull n8nio/n8n` Page: https://agentspot.dev/tools/n8n/ | Repo: https://github.com/n8n-io/n8n | Site: https://n8n.io
- **chrome-devtools-mcp** (Apache-2.0) — An official MCP server exposing Chrome DevTools to coding agents, letting them inspect, debug, and drive a real Chrome browser. [52,510 stars, 1,026,016 npm downloads/7d] | Can run locally | Install: `npm install chrome-devtools-mcp` Page: https://agentspot.dev/tools/chrome-devtools-mcp/ | Repo: https://github.com/ChromeDevTools/chrome-devtools-mcp | Site: https://npmjs.org/package/chrome-devtools-mcp
- **context7** (MIT) — An MCP server that feeds up-to-date, version-specific library documentation into LLMs and AI code editors so generated code targets current APIs. [62,353 stars, 354,384 npm downloads/7d] | Needs a hosted service | Install: `npm install @upstash/context7-mcp` Page: https://agentspot.dev/tools/context7/ | Repo: https://github.com/upstash/context7 | Site: https://context7.com
- **playwright-mcp** (Apache-2.0) — Microsoft's MCP server that gives agents structured, accessibility-tree-based control of a browser via Playwright, without relying on screenshots. [37,496 stars, 5,425,474 npm downloads/7d] | Can run locally | Install: `npm install @playwright/mcp` Page: https://agentspot.dev/tools/playwright-mcp/ | Repo: https://github.com/microsoft/playwright-mcp | Site: https://www.npmjs.com/package/@playwright/mcp
- **dify** (Other (see repo)) — A production-ready platform for building agentic LLM applications and workflows, combining a visual builder, RAG, and model orchestration. [156,951 stars, 248,872 docker downloads/7d] | Can run locally | Install: `docker pull langgenius/dify-api` Page: https://agentspot.dev/tools/dify/ | Repo: https://github.com/langgenius/dify | Site: https://dify.ai
- **fastmcp** (Apache-2.0) — A Pythonic framework for building Model Context Protocol servers and clients quickly, handling the protocol plumbing for you. [27,878 stars, 12,607,400 pypi downloads/7d] | Can run locally | Install: `pip install fastmcp` Page: https://agentspot.dev/tools/fastmcp/ | Repo: https://github.com/PrefectHQ/fastmcp | Site: https://gofastmcp.com
- **chroma** (Apache-2.0) — An open-source search and retrieval engine (vector database) for building AI applications with embeddings-based memory and RAG. [29,358 stars, 1,423,119 pypi downloads/7d] | Can run locally | Install: `pip install chromadb` Page: https://agentspot.dev/tools/chroma/ | Repo: https://github.com/chroma-core/chroma | Site: https://www.trychroma.com/
- **composio** (MIT) — A platform giving agents 1000+ tool integrations, tool search, managed authentication, and a sandboxed workbench for taking real-world actions. [30,293 stars, 355,465 pypi downloads/7d] | Needs a hosted service | Install: `pip install composio` Page: https://agentspot.dev/tools/composio/ | Repo: https://github.com/ComposioHQ/composio | Site: https://docs.composio.dev
- **OpenCLI** (Apache-2.0) — Command-line tool that turns dozens of websites and desktop apps into deterministic CLI commands, executed through your own logged-in browser — with an agent-skill layer so coding agents can drive it or author new adapters. [29,577 stars, 9,466 npm downloads/7d] | Can run locally | Install: `npm install @jackwener/opencli` Page: https://agentspot.dev/tools/opencli/ | Repo: https://github.com/jackwener/OpenCLI | Site: https://opencli.info/
- **E2B** (Apache-2.0) — Open-source secure cloud sandboxes that let AI agents run code and use real tools in isolated environments. [13,929 stars, 1,279,211 pypi downloads/7d] | Needs a hosted service | Install: `pip install e2b` Page: https://agentspot.dev/tools/e2b/ | Repo: https://github.com/e2b-dev/E2B | Site: https://e2b.dev/docs
- **Figma-Context-MCP** (MIT) — An MCP server that hands Figma layout and design data to AI coding agents so they can build UIs that match a design. [15,900 stars, 60,727 npm downloads/7d] | Needs a hosted service | Install: `npm install figma-developer-mcp` Page: https://agentspot.dev/tools/figma-context-mcp/ | Repo: https://github.com/GLips/Figma-Context-MCP | Site: https://www.framelink.ai/
- **TencentDB-Agent-Memory** (Other (see repo)) — TencentCloud's team-level memory hub for AI agents, distilling conversations, docs, and code into reusable, queryable memory assets that agents can share. [27,194 stars] | Needs a hosted service Page: https://agentspot.dev/tools/tencentdb-agent-memory/ | Repo: https://github.com/TencentCloud/TencentDB-Agent-Memory
- **github-mcp-server** (MIT) — GitHub's official MCP server, letting agents work with repositories, issues, pull requests, and Actions through the GitHub API. [33,138 stars] | Needs a hosted service Page: https://agentspot.dev/tools/github-mcp-server/ | Repo: https://github.com/github/github-mcp-server
- **Google Workspace CLI** (Apache-2.0) — Google's official command-line tool for Drive, Gmail, Calendar, Sheets, and the rest of Workspace, generated from the Discovery APIs — usable by humans and by agents needing programmatic Workspace access. [31,116 stars] | Needs a hosted service Page: https://agentspot.dev/tools/google-workspace-cli/ | Repo: https://github.com/googleworkspace/cli | Site: https://developers.google.com/workspace
- **firecrawl-mcp-server** (MIT) — The official Firecrawl MCP server, adding web scraping and search to MCP-compatible clients like Claude and Cursor. [7,503 stars, 22,552 npm downloads/7d] | Needs a hosted service | Install: `npm install firecrawl-mcp` Page: https://agentspot.dev/tools/firecrawl-mcp-server/ | Repo: https://github.com/firecrawl/firecrawl-mcp-server | Site: https://firecrawl.dev
- **mcp-toolbox** (Apache-2.0) — Google's open-source MCP server for databases, giving agents secure, pooled access to query many SQL and NoSQL data stores. [16,480 stars] | Can run locally Page: https://agentspot.dev/tools/mcp-toolbox/ | Repo: https://github.com/googleapis/mcp-toolbox | Site: https://mcp-toolbox.dev/documentation/introduction/
- **exa-mcp-server** (MIT) — An MCP server exposing Exa's neural web search and crawling so agents can retrieve current information from the web. [5,043 stars, 22,768 npm downloads/7d] | Needs a hosted service | Install: `npm install exa-mcp-server` Page: https://agentspot.dev/tools/exa-mcp-server/ | Repo: https://github.com/exa-labs/exa-mcp-server | Site: https://exa.ai/docs/reference/exa-mcp

## Coding

- **OpenCode** (MIT) — Open-source terminal coding agent built to work with any model provider — a fast, provider-agnostic harness with a polished terminal interface. [209,563 stars] | Can run locally Page: https://agentspot.dev/coding/opencode/ | Repo: https://github.com/anomalyco/opencode | Site: https://opencode.ai
- **mini-swe-agent** (MIT) — A deliberately minimal (~100-line) coding agent that resolves GitHub issues and assists at the command line, while scoring competitively on SWE-bench. [7,908 stars, 2,347,715 pypi downloads/7d] | Can run locally | Install: `pip install mini-swe-agent` Page: https://agentspot.dev/coding/mini-swe-agent/ | Repo: https://github.com/SWE-agent/mini-swe-agent | Site: https://mini-swe-agent.com
- **oh-my-pi** (MIT) — Terminal coding agent with hash-anchored edits, language-server integration, browser control, Python execution, and subagents in an optimized tool harness. [32,915 stars] Page: https://agentspot.dev/coding/oh-my-pi/ | Repo: https://github.com/can1357/oh-my-pi | Site: https://omp.sh
- **CodeWhale** (MIT) — A Rust terminal coding agent that runs agentic workflows — file edits, shell commands, git, MCP tools, and sub-agents — against many model providers, from DeepSeek and Claude to local Ollama. [41,029 stars] | Can run locally Page: https://agentspot.dev/coding/codewhale/ | Repo: https://github.com/Hmbown/CodeWhale | Site: https://codewhale.net/
- **AionUi** (Apache-2.0) — Desktop cowork app that runs and orchestrates twenty-plus CLI coding agents — Claude Code, Codex, OpenCode and others — behind one shared graphical workspace. [33,058 stars] Page: https://agentspot.dev/coding/aionui/ | Repo: https://github.com/iOfficeAI/AionUi | Site: https://www.aionui.com
- **Claude Code** (Other (see repo)) — Anthropic's agentic coding tool: an agent that lives in the terminal (and IDEs), reads and edits codebases, runs commands, and automates development workflows. [147,733 stars, 283 docker downloads/7d] | Needs a hosted service | Install: `docker pull anthropics/claude-code` Page: https://agentspot.dev/coding/claude-code/ | Repo: https://github.com/anthropics/claude-code | Site: https://code.claude.com/docs/en/overview
- **Codex CLI** (Apache-2.0) — OpenAI's open-source coding agent for the terminal: a Rust CLI that pairs OpenAI models with local code reading, editing, and command execution. [126,118 stars] Page: https://agentspot.dev/coding/codex/ | Repo: https://github.com/openai/codex
- **Cursor** (Other (see repo)) — The AI-native code editor: a commercial IDE with built-in agents that read, edit, and refactor across a codebase. [33,250 stars] | Needs a hosted service Page: https://agentspot.dev/coding/cursor/ | Repo: https://github.com/cursor/cursor | Site: https://cursor.com
- **gemini-cli** (Apache-2.0) — Google's open-source AI agent that brings Gemini into the terminal for coding, automation, and MCP-connected workflows. [107,138 stars, 289,730 npm downloads/7d] | Needs a hosted service | Install: `npm install @google/gemini-cli` Page: https://agentspot.dev/coding/gemini-cli/ | Repo: https://github.com/google-gemini/gemini-cli | Site: https://geminicli.com
- **qwen-code** (Apache-2.0) — The Qwen team's open-source coding agent for the terminal, pairing Qwen's coder models with an agentic command-line workflow. [28,077 stars, 56,787 npm downloads/7d] | Can run locally | Install: `npm install @qwen-code/qwen-code` Page: https://agentspot.dev/coding/qwen-code/ | Repo: https://github.com/QwenLM/qwen-code | Site: https://qwenlm.github.io/qwen-code-docs/en/users/overview

## Learn

- **ai-agent-book** (Apache-2.0) — A complete open-source book on agent engineering — 'Understanding AI Agents in Depth: Design Principles and Engineering Practice' by Li Bojie — publishing the full text, compiled PDF, and per-chapter companion code. [50,287 stars] Page: https://agentspot.dev/learn/ai-agent-book/ | Repo: https://github.com/bojieli/ai-agent-book
- **generative-ai-for-beginners** (MIT) — Microsoft's 21-lesson course teaching the fundamentals of building generative-AI applications, from prompting to building agents. [120,351 stars] Page: https://agentspot.dev/learn/generative-ai-for-beginners/ | Repo: https://github.com/microsoft/generative-ai-for-beginners
- **learn-claude-code** (MIT) — A from-scratch tutorial that builds a nano Claude Code-style agent harness — 'Bash is all you need' — teaching the agent loop by implementing one. [77,494 stars] Page: https://agentspot.dev/learn/learn-claude-code/ | Repo: https://github.com/shareAI-lab/learn-claude-code | Site: https://learn.shareai.run
- **AI Agents for Beginners** (MIT) — Microsoft's 18-lesson course on building AI agents, each lesson pairing a written explanation with runnable notebooks — covering agentic design patterns, tool use, planning, multi-agent systems, and production concerns. [75,496 stars] Page: https://agentspot.dev/learn/ai-agents-for-beginners/ | Repo: https://github.com/microsoft/ai-agents-for-beginners | Site: https://aka.ms/ai-agents-beginners
- **Hugging Face Agents Course** (Apache-2.0) — Hugging Face's free course on agents, running from first principles through building and evaluating your own — with hands-on units, a certification track, and coverage of several frameworks rather than one vendor's. [32,802 stars] Page: https://agentspot.dev/learn/huggingface-agents-course/ | Repo: https://github.com/huggingface/agents-course | Site: https://huggingface.co/learn/agents-course
- **GenAI Agents** (Other (see repo)) — A large collection of agent tutorials, each a self-contained notebook implementing one technique — from a basic conversational agent up through multi-agent orchestration, memory, and evaluation. [24,366 stars] Page: https://agentspot.dev/learn/genai-agents/ | Repo: https://github.com/NirDiamant/GenAI_Agents | Site: https://diamant-ai.com
- **Anthropic Courses** (Other (see repo)) — Anthropic's own teaching material as runnable notebooks — prompt engineering, tool use, retrieval, and evaluation — written against the Claude API but largely transferable to other models. [22,857 stars] Page: https://agentspot.dev/learn/anthropic-courses/ | Repo: https://github.com/anthropics/courses
- **Agents Towards Production** (Other (see repo)) — Code-first tutorials on the parts of agent work that come after the demo: deployment, observability, evaluation, cost control, and security — organised as independent chapters you can take in any order. [21,487 stars] Page: https://agentspot.dev/learn/agents-towards-production/ | Repo: https://github.com/NirDiamant/agents-towards-production | Site: https://diamant-ai.com