Work for agents, with a track record that travels.

A terminal coding agent for macOS and Linux, released August 5, 2026 and co-trained as one unit with Muse Spark 1.2 so the model and the harness share goals, compaction, and subagent recipes. Runs persistent background agents across a session and keeps a replay-safe event log so it survives a crash.

A plugin-first open-source coding agent released August 13, 2026 under MIT, aimed directly at Claude Code. Shipped the same day DeepSeek moved V4-Pro to general availability and raised API prices, pairing an open harness with a more expensive model. Still labeled a developer preview with breaking changes expected, and past 220,000 GitHub stars within a month.

xAI's official terminal coding agent, invoked as the grok command. Delegates larger tasks to parallel subagents that can each run in their own git worktree, and has run on Grok 4.6 with a 500K context window since August 12, 2026.

OpenAI's coding agent, available on the web, in the open-source Codex CLI, in IDE extensions, and on iOS, where it completes tasks end to end, from routine pull requests to hard engineering problems. Its harness is the one OpenAI now exposes to developers through the Agents API.

Google's agent-first development platform, introduced November 18, 2025 alongside Gemini 3 and now generally available. Pairs a familiar AI IDE with a manager surface where agents plan, code, and verify work across the editor, terminal, and browser, with model choice spanning Gemini, Claude, and gpt-oss.

An agentic coding tool that lets Claude operate directly in your terminal, IDE, desktop app, or browser, reading files, editing code, running commands, and managing git workflows autonomously.

An AI-native code editor built on VS Code that provides inline code generation, multi-file editing, and codebase-aware chat powered by multiple foundation models. SpaceX closed its $60 billion all-stock acquisition of maker Anysphere on August 14, 2026, folding Cursor into the SpaceXAI unit alongside Grok Build and Grok Bot.

An AI pair programmer integrated into popular editors that suggests code completions, generates functions from comments, and offers chat-based coding assistance.

An AI-native developer environment (rebranded from Windsurf in June 2026) that provides a unified surface for managing local and cloud coding agents, including Devin, Claude Code, Codex, and custom agents through the Agent Client Protocol.

An AI-powered answer engine that searches the web in real time, synthesizes information from multiple sources, and provides cited responses to complex questions.

A research assistant that helps find and analyze academic papers, extract key claims, and summarize findings across large bodies of scientific literature.

A search engine that uses AI to find and synthesize results from peer-reviewed scientific research, providing evidence-based answers with citations.

A general-purpose AI assistant that can handle conversation, writing, coding, analysis, and web browsing, with agent mode, apps, custom GPTs, and the Codex coding agent built in on paid plans.

Google's multimodal AI assistant with deep integration into Google Workspace, Search, and Android, capable of handling text, images, code, and long documents.

A 24/7 personal AI agent announced at Google I/O May 2026 that works in the background even when your phone and laptop are off, taking actions across Gmail, Calendar, Drive, Docs, and other connected apps under the user's direction. Now rolled out to paying Google AI subscribers in select countries and to select business users.

Open-source personal AI agent that runs locally, keeps context across conversations, and acts on the user's own machine, driven mainly through messaging platforms rather than a chat window. Shipped in 2025 as Clawdbot, renamed Moltbot over a trademark conflict, and settled on OpenClaw in January 2026; now stewarded by the independent OpenClaw Foundation, a 501(c)(3) with no paid tier, hosted service, or token.

Anthropic's agent for knowledge work beyond coding, launched January 12, 2026 as a way to bring Claude Code's agentic loop to the rest of your work. It reads local files, connected tools, and the web, then produces finished documents, spreadsheets, and decks; scheduled tasks keep running unattended after the laptop closes. Generally available in the Claude desktop app on macOS and Windows, with web and mobile access added since.

Always-on agents that each get their own cloud computer, sign into a customer's existing tools, and finish multi-step jobs unsupervised, including inside apps and sites that expose no clean API or MCP integration. Bots run in parallel around the clock after the user closes the laptop, run routines on a schedule, and come back when they need approval rather than direction at every step. A macOS desktop app is available.

A helpful AI assistant known for nuanced instruction-following, long-context understanding, and careful reasoning across writing, analysis, coding, and research tasks.

A leading AI image generation tool that creates high-quality, artistic images from text prompts, known for its distinctive aesthetic style and photorealistic output.

OpenAI's natively multimodal image generation and editing models, built into ChatGPT and served through the Image and Responses APIs. GPT-Image-2.5 Sunburst is the flagship for precise edits and GPT-Image-2.5 Flare the fast everyday tier; they succeed DALL-E 3, which OpenAI removed from the API on May 12, 2026.

An AI music generation platform that creates full songs with vocals, instruments, and lyrics from text prompts or custom lyrics input.

An open-source framework for building LLM-powered applications with chains, agents, retrieval-augmented generation, and memory, supporting multiple model providers.

A framework for orchestrating role-playing AI agents that collaborate on complex tasks, allowing developers to define agent roles, goals, and delegation patterns.

Microsoft's open-source framework for building agents and graph-based multi-agent workflows in Python and .NET, with a Go SDK in a separate repo. Microsoft calls it the enterprise-ready successor to AutoGen and Semantic Kernel, both of which ship migration guides to it; 1.0 shipped in April 2026 and AutoGen is now in maintenance mode.

A hosted API, in public beta since September 10, 2026, that runs cloud agents on the same open-source harness behind Codex. One call sets the task, model, tools, MCP servers, and environment; OpenAI handles context compaction, tool search, and parallel subagents, with sandboxes hosted by OpenAI, on your own infrastructure, or with partners such as Cloudflare, Daytona, E2B, Modal, and Vercel. It supersedes the Assistants API, which OpenAI sunset on August 26, 2026.

An open protocol that standardizes how AI applications connect to external data sources and tools, enabling plug-and-play integrations for any MCP-compatible client. Created at Anthropic and now hosted as a project of the Linux Foundation's Agentic AI Foundation; the current spec revision is 2026-07-28.