An agent is more than its model
Tools, connectors, execution environments, memory, and permissions determine what an agent can actually do.
THE AI LANDSCAPE 07 / APPLICATIONS & WORKFLOWS
Compare selected agents by the work they handle, where they operate, and their current release status. Product boundaries and access are changing quickly.
THE BIG PICTURE
Selected coverage. Check each entry for context and evidence.
Try a different name or a broader category.
A manually reviewed snapshot as of September 23, 2026. Colors organize the evidence; they do not score quality. On smaller screens, swipe across the comparison.
CONTEXT MAKES THE DIFFERENCE
Our interpretation of the comparison.
Tools, connectors, execution environments, memory, and permissions determine what an agent can actually do.
Unified products can replace older standalone names. A current product map needs to follow those transitions, not just model releases.
A preview endpoint, a paid desktop product, and a regional rollout are different ways to access an agent. Read the release notes before choosing.
TRANSPARENT BY DESIGN
Official announcements, product documentation, and model repositories. All linked sources were checked on September 23, 2026. Publication dates and review dates are shown separately.
Selected end-user and developer agents with official product pages or recent release notes. Work-focus labels describe vendor-supported use cases, not independently benchmarked reliability. We separate coding-focused agents from broader work and personal agents; the groups overlap. Access notes are a dated snapshot and do not guarantee eligibility, quota, geography, or feature availability for every account. Product permissions remain configurable and should be reviewed in the product.
Dates on continuously updated documentation are labeled “Current documentation” unless a specific publication date is available. This page is maintained manually; the review date is not a live-feed timestamp. Company announcements describe the company’s claims and plans, not an independent audit.
OpenAI / Work & personal
ChatGPT Work brings tasks and work context together. Consult the current product guide for the boundary between Work and Codex and for supported access.
OpenAI / Coding agents
Codex is the coding-focused product in OpenAI’s Work and Codex guide. Its ability to act depends on the connected development environment and tools.
Anthropic / Work & personal
Anthropic is merging Cowork and chat into Claude over a staged rollout. The announcement starts with Pro and Max, with other account types following; it should not be described as complete for everyone.
Anthropic / Coding agents
Claude Code is Anthropic’s coding agent for working with a codebase and development tools. Its product surface is distinct from the broader Claude work experience.
SpaceXAI / Work & personal
Grok Bot runs work on its own cloud computer and retains context across tasks. The current launch page still labels the product beta and lists supported Grok and Cursor subscriptions.
Meta / Work & personal
Meta’s September 8 announcement describes a US rollout. Muse operates in its own virtual machine and asks for approval before sensitive actions such as sending an email or making a purchase.
Google / Coding agents
The Gemini API changelog introduces antigravity-preview-09-2026 on September 17. It replaces an earlier preview; integrations need to follow the endpoint lifecycle.
Kiro / AWS / Coding agents
Kiro organizes agentic development around specifications and project context. Its IDE, CLI, and web experiences are related surfaces rather than a single uniform feature set.
Mistral AI / Work & personal
Mistral’s September 22 release notes describe the merger of Vibe Chat and Work, with migration beginning September 14 for Free, Pro, and Teams. Enterprise migration is scheduled to start October 1, after this page’s review date.