AI Agents: 6 Major Platforms Compared — Claude Cowork vs Microsoft vs ChatGPT vs Cursor

AI Tools Insight • 2026-10-27 • AI Agents Claude Cowork Microsoft Agent 365 ChatGPT Cursor Gemini Perplexity Comparison

AI agents became a product you can buy. Claude Cowork works on your files. Microsoft puts an agent inside Outlook and Teams. Cursor runs coding agents in parallel. Perplexity Computer takes a goal and returns a document.

The field also moves. One product named in the May 2026 version of this page no longer exists. This revision states only what a vendor page confirms.

Checked against vendor pages on 2026-10-27. Where a vendor shows only an annual-billing rate, this page shows that rate and says so.

Quick Comparison Table

Platform Entry Paid Runs On Local Files? Best For
Claude Cowork $20/mo Pro ($17 annual) Desktop, web, mobile, Chrome Yes, on desktop Document and file work
Microsoft Copilot Cowork $30.00/user/mo yearly M365 apps and cloud Through OneDrive and SharePoint M365 enterprises
ChatGPT Work $8/mo Go Web, desktop, mobile No General tasks and research
Cursor $8/mo Plus Desktop IDE Yes, your codebase Coding
Google Gemini $19.99/mo AI Pro Web and mobile No Workspace work
Perplexity Computer $17/mo Pro (annual billing) Web and desktop app No Cited research

Two entries changed since May. ChatGPT agent is discontinued, and ChatGPT Work replaces it. Cursor no longer has a $20 Pro tier.

Claude Cowork (Anthropic) — Best for Document and File Work

Pricing: Free $0 → Pro $20/mo billed monthly or $17/mo with an annual subscription ($200 up front) → Max Max 5x from $100/mo and Max 20x at $200/mo

Claude Cowork brings the agent loop of Claude Code to knowledge work. You describe the outcome, then step away. You return to finished documents, organized files, and synthesized research.

How it runs: Cowork runs sessions remotely in the cloud. On the desktop app, Claude reads and writes your local files without manual uploads. The desktop app also carries a built-in browser, so Claude can open pages, click, and fill forms with nothing to install.

Availability: Cowork runs on all paid plans. The desktop app covers macOS, Windows, and Linux (beta). Web and Claude Mobile cover Pro, Max, and Team (beta on Team), and Enterprise when an owner enables it. The Claude in Chrome side panel covers all paid plans.

Models: Claude Opus 5.5, Sonnet 5.5, and Haiku 5.5 are the current family. Max 5x gives 5 times the Pro usage per 5-hour session, and Max 20x gives 20 times.

Strengths:

Weaknesses:

Best for: Document-heavy work, file organization, and research synthesis.

Sources: anthropic.com/pricing, support.claude.com Cowork, docs.anthropic.com models

Microsoft Copilot Cowork and Agent 365 — Best for M365 Enterprises

Pricing: Microsoft 365 Copilot add-on $30.00/user/mo paid yearly or $31.50 paid monthly → Microsoft 365 E7 $99.00/user/mo paid yearly

Copilot Cowork is the agent. You describe the task, and it does the work across your Microsoft 365 environment. It performs these actions:

You approve each action before it happens.

Copilot Cowork ships with many models. An admin can turn off the Anthropic model family in the Microsoft 365 admin center. That setting applies to the whole tenant.

Microsoft Agent 365 is a different product. It governs agents. It does not do the work. E7 bundles it with Microsoft 365 E5, Microsoft 365 Copilot, and Microsoft Entra Suite.

Strengths:

Weaknesses:

Best for: Organizations on Microsoft 365, especially teams that need an approval step on every action.

Sources: microsoft.com E7, learn.microsoft.com Cowork, Cowork models

ChatGPT Work (OpenAI) — Best for General Tasks

Pricing: Free $0 → Go $8/mo → Plus $20/mo → Pro $200/mo

ChatGPT agent is no longer available. OpenAI states that ChatGPT Work replaces it for longer tasks and finished deliverables. A cloud browser covers supported browser workflows. A page that still describes agent mode sends readers to a product they cannot buy.

The model depends on the plan. GPT-6.1 Sol is not available on Free or Go. Plus includes it. Pro adds an expanded allowance, and GPT-6 Pro runs on GPT-6 Astra.

Strengths:

Weaknesses:

Best for: General tasks, web research, and written deliverables.

Sources: openai.com/chatgpt/pricing, help.openai.com ChatGPT agent, ChatGPT Go

Cursor — Best for Coding

Pricing: Free $0 → Plus $8/mo → Premium $30/mo → Super $100/mo → Ultra $200/mo → Teams $30+/user/mo

Cursor reorganised its plans. The former $20 Pro tier is gone. Plus costs $8/mo. Premium costs $30, Super costs $100, and Ultra costs $200.

Strengths:

Weaknesses:

Best for: Software developers who want agents that read, write, and refactor code.

Source: cursor.com/pricing

Google Gemini — Best for Google Ecosystem

Pricing: Free $0 → AI Pro $19.99/mo → AI Ultra from $99.99/mo

Gemini agent work stays inside Google Workspace. It drafts email, writes documents, analyses spreadsheets, and summarises meetings inside Google's apps.

Gemini 4 Argon is the current flagship on Google's model page. The Gemini API documentation states a unified 1M context window for it.

Strengths:

Weaknesses:

Best for: Teams already on Google Workspace, and long-document analysis.

Sources: one.google.com/about/ai-premium, deepmind.google/models/gemini, ai.google.dev models

Perplexity Computer — Best for Cited Research

Pricing: Free (limited) → Pro $17/mo → Max $167/mo. Both paid rates are the annual-billing figures on the Perplexity pricing page. The page publishes no monthly rate.

Perplexity announced Computer on 26 February 2026. It takes a high-level objective and returns apps, documents, and tools.

The model count needs care. The changelog states that Computer orchestrates work across 19 models in parallel. The current pricing page names six of them: GPT-6.1 Sol, Gemini 3.8 Flash, Claude Sonnet 5.5, Kimi K3, GLM 5.3, and Grok 4.7.

Pro adds expanded Computer access, 4,000 bonus credits, deep research, and access to the top models. Max adds maximum Computer usage, 10,000 monthly credits, and 35,000 bonus credits.

Strengths:

Weaknesses:

Best for: Research, competitive analysis, and cited reports.

Sources: perplexity.ai/pro, Perplexity changelog 2026-02-26

Feature Comparison Matrix

Feature Claude Cowork Copilot Cowork ChatGPT Work Cursor Gemini Perplexity
Local file system Yes, on desktop Via OneDrive No Yes, codebase No No
Desktop app Yes No Yes Yes, IDE No Yes
Web access Yes Yes Yes No Yes Yes
Mobile Yes, Pro and up Yes, M365 app Yes No Yes Yes
Built-in browser Yes, on desktop Not stated Cloud browser No Not stated Yes
Approval per action Not stated Yes Not stated No Not stated No
Enterprise identity Team and Enterprise Entra Enterprise Teams Workspace Not stated
Free tier Yes No, add-on only Yes, GPT-6 Yes Yes Yes
Published monthly price Yes Yes Yes Yes Yes No

Cells marked "Not stated" mean that no vendor page makes the claim.

Pricing Breakdown by Platform

Tier Claude Microsoft ChatGPT Cursor Gemini Perplexity
Free $0 No free Cowork $0, GPT-6 $0 $0 $0, limited
Under $20/mo $17 annual No $8 Go $8 Plus No $17 annual
$20-30/mo $20 monthly $30.00 yearly $20 Plus $30 Premium $19.99 AI Pro No published rate
$100/mo Max 5x No No $100 Super Ultra from $99.99 No
$200/mo Max 20x No $200 Pro $200 Ultra No $167 Max, annual
Enterprise Quoted E7 $99.00 yearly Quoted Quoted Quoted Personal use only

Quick Decision Guide

If you need... Choose Entry price
An agent that edits your own files Claude Cowork (Pro) $20/mo
Governance for agents on Microsoft Microsoft E7 $99.00/user/mo
An agent inside Outlook and Teams Copilot Cowork $30.00/user/mo
Browser and general tasks ChatGPT Work (Plus) $20/mo
Coding agents Cursor (Plus) $8/mo
Google Workspace work Gemini (AI Pro) $19.99/mo
Research with citations Perplexity (Pro) $17/mo annual

Summary

The market split into two shapes. Desktop agents reach your files. Claude Cowork is the clearest example. Cloud agents run on vendor infrastructure and work with your SaaS applications. Copilot Cowork, ChatGPT Work, Gemini, and Perplexity Computer fit that shape. Cursor sits apart as an agent inside a code editor.

For one person who needs local file access, Claude Cowork Pro at $20/mo covers the widest set of surfaces on one plan.

For a Microsoft estate, start with Copilot Cowork at $30.00/user/mo paid yearly. Move to E7 at $99.00/user/mo when you need Agent 365 governance across many agents.

For coding, Cursor Plus at $8/mo is the cheapest entry. The useful tier is probably Premium at $30/mo.

Every price on this page was checked against the vendor's own page on 2026-10-27. Rates change. Check each page before you buy.

What changed in this revision

Claim in the May version What the vendor pages show now
ChatGPT agent on Plus, drives a browser ChatGPT agent is no longer available. ChatGPT Work and a cloud browser replace it
Cursor Pro $20/mo and Ultra $200/mo Plus $8, Premium $30, Super $100, Ultra $200. No $20 tier
Microsoft Agent 365 standalone at $15/user/mo No vendor page states this. Agent 365 arrives in E7 at $99.00/user/mo
Claude Opus 4.6 and Sonnet 4.5 The current family is Opus 5.5, Sonnet 5.5, and Haiku 5.5
Claude Cowork is desktop only Cowork also runs on web, mobile, and a Chrome panel
Claude Cowork works on your machine Cowork runs sessions in the cloud. The desktop app provides local file access
Perplexity Pro $20/mo and Max $200/mo $17/mo and $167/mo on annual billing. No monthly rate published
Gemini 3.1 Pro, 94.3% GPQA Diamond Gemini 4 Argon is the flagship. No vendor page states the figure
AI Ultra $99.99 (5x) and $199.99 (20x) "From $99.99/mo" is the only Ultra rate
Claude Opus 4.6 topped a Palisade benchmark at 81% No vendor page states this. Removed
Gartner: 40% of enterprise apps will include agents No vendor page states this. Removed

Two claims survived the challenge. The May version said Computer coordinates 19 models. A first draft of this revision removed that figure as unsourced. The Perplexity changelog does state 19 models in parallel, so the figure is back with its source. The Max 5x and Max 20x tiers are also correct.

Try Claude Free →

Try Claude Free

Anthropic's AI assistant with Cowork agent. Start free.

Get Started — from $0/mo
Share this article
AI Tools Insight may earn a commission from some links. Editorial independence is always maintained.