AI coding agents, SDKs and agent tooling

A web scraping and crawling API (and open-source engine) that turns sites into LLM-ready data.

Google's open-source AI agent that brings Gemini into the terminal.

OpenHands: an open-source platform for AI-driven software development.

HeyGen's open-source framework that renders HTML, CSS and seekable animations to MP4 video, with agent skills and a Claude Code plugin.

An autonomous coding agent available as an IDE extension, CLI or SDK.

Reference repository of Claude Code practices for subagents, commands, skills, hooks, MCP, settings and memory, with example implementations.

A framework for orchestrating role-playing, autonomous AI agents.

Web app, desktop app and MCP server for creating and editing draw.io diagrams by chatting with an AI model; exports .drawio, .png or .svg.

Trail-code (formerly graft): local code index and team-learnings layer that gives Claude Code, Cursor, Codex and other agents codebase context; optional Trail workspace.

An open-source coding agent (formerly sst/opencode).

goose: an open-source, extensible AI agent that installs, executes, edits and tests code with any LLM.

LangChain's framework for building resilient, stateful agents.

Open-source AI memory platform that turns documents and conversations into a searchable knowledge graph for agents.

Zep's open-source Python framework for building temporal knowledge graphs as agent memory; the repo also includes an MCP server.

A Python framework from Prefect for building MCP servers, clients and interactive apps; a TypeScript counterpart is maintained separately.

TypeScript framework for building, testing and shipping MCP servers and MCP Apps, with a built-in Inspector; made by Manufact, which offers hosted deploys.

Local, browser-based review surface where you annotate coding-agent plans, specs and code diffs and send the feedback back to the agent via hooks.

designlang: CLI that extracts a website's design system into DTCG tokens, Tailwind, Figma variables and more, with an MCP server for coding agents.

The official Python SDK for the Model Context Protocol, for building MCP servers and clients over stdio, Streamable HTTP and SSE.

NVIDIA's security scanner for agent skills: static analysis plus optional LLM review for prompt injection, data exfiltration and other risks.

OpenAI's lightweight coding agent that runs in your terminal.

Microsoft's open-source Model Context Protocol curriculum, with lessons and examples in .NET, Java, TypeScript, JavaScript, Rust and Python.

GitHub CLI extension that compiles Markdown agentic workflows into GitHub Actions, running engines such as Copilot, Claude Code or Codex in sandboxed jobs.

Hook that blocks destructive Git and file-system commands and access to secrets before a coding agent runs a tool call; per its README, it is not a sandbox.

A CLI for analyzing coding-agent token usage and cost from local logs.

The official TypeScript SDK for the Model Context Protocol, for building MCP servers and clients; v2 is the stable line per its README.

An open-source coding agent from Continue.

gws: a CLI for Google Workspace APIs (Drive, Gmail, Calendar and more) with JSON output and agent skills; per its README, not an official Google product.

The Python SDK for building agents on top of Claude Code.

Semantic memory for coding agents, kept as Markdown files with a Milvus index; plugins for Claude Code, Codex, OpenCode and others, plus a CLI and Python API.

Hivemind by Activeloop: cloud-backed shared memory that turns agent session traces into reusable skills across Claude Code, Codex, Cursor and others.

AGENTS.md: a simple, open format for giving coding agents project instructions.

Microsoft's Python library and CLI that converts PDF, Office, HTML, images and more to Markdown for LLMs; includes a markitdown-mcp package.

A visual testing tool for MCP servers.

Anthropic's agentic coding tool for the terminal, IDEs and GitHub.

A GitHub Action that runs Claude Code on pull requests and issues.

Testing and evals platform for MCP servers and apps: playground, OAuth debugger, JSON-RPC traces and CLI. Mostly Apache-2.0; evals code licensed separately.

JetBrains' official TeamCity CLI for builds, logs and queues, with agent-skills install support.

AI pair programming in your terminal.

A Claude Code guide covering workflows, hooks, skills and MCP servers, with templates and a stdio MCP server for querying it; guide licensed CC BY-SA 4.0.

Indexes past Claude Code, Codex, pi and Hermes sessions into durable searchable memory publishable on Hugging Face Hub.

NVIDIA's framework for evaluating agent skills: static quality checks, overlap detection, synthetic eval data and live runs with and without a skill.

The TypeScript SDK for building agents on top of Claude Code.