Run programmable browser sessions and web actions for AI agents and applications.
Project overview
Steel exposes a browser API that manages Chrome sessions, pages, processes, cookies, local storage, proxies, extensions, and cleanup for AI agents or automation applications. Developers can connect with Puppeteer, Playwright, or Selenium, or call quick endpoints for scraping pages, taking screenshots, and producing PDFs. It can be used through Steel Cloud or deployed locally with Docker and other supported setups. The project is still described as a public beta.
Modify code across files and run tests from natural-language terminal instructions.
Project overview
Codebuff is an open-source terminal coding assistant that coordinates specialized agents to understand a project, find the right files, make cross-codebase changes, and run tests. Its SDK can embed agents in applications, CI/CD systems, or custom workflows, with model selection available through OpenRouter. A separate ad-supported Freebuff edition removes subscriptions and credits, but its models, daily sessions, and full-access mode vary by region.
Publish, discover, and verify community MCP server listings.
Project overview
MCP Registry exposes a searchable list of MCP servers to clients and provides an API, publisher tooling, and a locally runnable service for maintainers. Publishing verifies namespace ownership through GitHub OAuth, GitHub OIDC, DNS, or HTTP challenges. The documented service remains in preview, with the v0.1 API temporarily frozen for integration validation rather than declared generally available.
Build controllable AI agents in Python or TypeScript
Project overview
Strands Agents is a monorepo with Python and TypeScript SDKs for building model-driven agents. The SDKs support tools, MCP, streaming, multi-agent patterns, hooks, guardrails, context management, and multiple model providers, with examples for local development and deployment. Python requires 3.10 or newer and the TypeScript package requires Node.js 20 or newer; the default quickstart uses Amazon Bedrock credentials.
Let AI agents control browsers through sandboxed JavaScript.
Project overview
Launch Chromium or attach to an existing Chrome session, then navigate, click, fill forms, inspect pages, and capture screenshots through the Playwright API while keeping pages alive across scripts. Dev Browser runs code in a QuickJS WASM sandbox without host access and restricts file operations to its own temporary directory, making it suitable for browser automation by coding agents.
Let MCP-compatible AI agents control Windows applications, UI elements, keyboard, and mouse.
Project overview
Windows-MCP exposes Windows computer-use capabilities to MCP-compatible agents. Its tools cover clicking, typing, scrolling, dragging, shortcuts, screenshots, UI and display state, application and window control, and other system actions without requiring a particular vision model. It can run through uvx over stdio or an explicitly bound HTTP transport. Because the server operates with full system access and can perform irreversible actions, deployments should use authentication, IP and tool allowlists, and other restrictions; anonymized telemetry can be disabled.
Manage OpenCode coding sessions from desktop, web, PWA, or VS Code.
Project overview
OpenChamber is a visual interface for OpenCode that presents chats, diffs, files, terminals, tool progress, and Git or GitHub workflows across desktop, browser, PWA, and VS Code surfaces. It can run multiple agents in isolated worktrees, continue sessions from another device, and connect through tunnels or SSH to remote instances. The desktop package bundles a matching OpenCode CLI, while the web, CLI, and VS Code editions depend on an existing OpenCode installation.
Desktop appAI and agent developmentAgent Client Protocol
deepchat
@ThinkInAIXYZ·TypeScript
Run models, Skills, MCP tools, and resumable agent sessions in one desktop client.
Project overview
DeepChat is a local-first desktop client for running cloud and local models alongside agent tools. It supports installable Skills, MCP services, and ACP-compatible agents as first-class session options, while its Tape and Trace views preserve context, tool calls, requests, and results for inspection or recovery. Users can also control sessions through messaging integrations such as Telegram, Feishu/Lark, Discord, and others, making it suited to longer-running agent workflows.
Collection or directorySecurity and privacyClaude Code
skills
@trailofbits·Python
Install security auditing, vulnerability research, and development skills for Claude Code or Codex.
Project overview
Trail of Bits Skills Marketplace packages reusable plugins for AI-assisted security analysis, vulnerability research, testing, and selected development workflows. Users can add the marketplace to Claude Code and install individual plugins through its menu, while Codex can load the same marketplace through Claude plugin compatibility. The collection is aimed at security researchers and engineering teams that want repeatable audit procedures available to their coding agents.
ConnectorAI and agent developmentModel Context Protocol
klavis
@Klavis-AI·Python
Connect AI agents to prebuilt MCP integrations with OAuth and per-user instances.
Project overview
Klavis gives developers SDK and API building blocks for connecting agents to external services through MCP. Applications can create per-user server instances, combine services such as Gmail and Slack behind a Strata endpoint, or use individual integrations from a catalog of more than one hundred options with OAuth support. Open-source components can also run locally through Docker or a pipx-installed Strata package. Calls to the hosted Klavis service require an API key.
Install a focused skill set for Vue, Vite, and Nuxt development.
Project overview
This collection combines Anthony Fu’s opinionated project and design conventions with skills generated from official Vue, Nuxt, Vite, Vitest, UnoCSS, pnpm, and related documentation. Vendored skills extend the set to adjacent tools, while git submodules help track upstream sources. It targets modern Vue/Vite work and remains a proof of concept that has not been fully tested in practice.
Orchestration frameworkAI and agent developmentModel Context Protocol
ruoyi-ai
@ageerle·Java
Build enterprise AI agents with visual workflows, RAG, and multiple models.
Project overview
RuoYi AI is a full-stack platform for enterprise agent applications. It connects multiple model providers, builds local RAG knowledge bases, and lets teams compose model calls, email, human review, and other steps in a visual workflow. Its agent layer supports multiple decision modes, multi-agent coordination, MCP, and Skills extensions. A Docker Compose setup can start the backend, administration UI, user UI, and supporting services together.
CLI appResearch and knowledgeModel Context Protocol
notebooklm-mcp-cli
@jacob-bd·Python
Manage and query Google NotebookLM through a CLI or MCP server.
Project overview
Install one Python package to get both a command-line client and an MCP server for NotebookLM. You can create notebooks, add URL, text, Drive, or file sources, run queries, start web or Drive research, generate Studio content, and download resulting artifacts. Batch operations, cross-notebook queries, sharing, and Drive-source synchronization make it useful for scripts and AI assistants that need more than the web interface.
Build and self-host LLM agent backends behind a unified API.
Project overview
TaskingAI is a backend platform for developing and deploying LLM-based agents. It brings models from multiple providers, tools, RAG systems, assistants, and conversation history under one console and unified API. Developers can prototype configurations in the UI, then integrate them into multi-tenant products through REST endpoints or the Python SDK. The community edition can be self-hosted with Docker Compose and requires configuration for the selected model services.
Handles interactive or automated terminal development with multiple models, subagents, MCP, and ACP.
Project overview
Kode is an open-source terminal assistant with multi-model switching, expert consultation, specialized subagents, code editing, command execution, MCP extensions, AGENTS.md discovery, and an ACP server mode. It supports interactive, non-interactive, and plan workflows, plus plugins and model configuration files. Its default mode skips most approval prompts, so safe mode is better for sensitive work. The repository also says a major maintenance and refactoring phase runs from July through December 2026.
Run parallel coding agents in isolated Git worktrees
Project overview
Emdash is a desktop environment for running coding agents in parallel, with each task isolated in its own Git worktree for reviewing diffs, CI checks, pull requests, and merges. It supports local projects and SSH/SFTP remote machines, and keeps app state in local SQLite. Emdash does not upload code or chats to its servers, but the selected agent CLIs may still send content to their own model providers.
Coordinate multiple LLM agents for task solving and behavioral simulations.
Project overview
AgentVerse offers separate frameworks for collaborative task solving and for simulating interactions among multiple LLM-based agents. Its examples cover software teams, classroom environments, games, and other research scenarios, with support for hosted APIs and selected local-model setups. The simulation code is being refactored, and the project directs users who need a stable simulation-only version to the release-0.1 branch.
Manage organizational knowledge, AI agents, and workflows in one controlled collaboration platform.
Project overview
Magic is an AI collaboration platform for solo operators, small teams, and enterprises, bringing personal assistants, specialist agents, internal systems, and shared projects into one workspace. It can turn agent output into reports, dashboards, presentations, or spreadsheets while adding per-department, per-user, or per-agent budgets, approval gates for risky actions, and sandbox isolation. Users can choose the hosted service or self-host with Docker on macOS or Linux; Windows self-hosting is still planned.