Anthropic
Claude Code / PDF analysis
The tools reference and the Claude Code changelog both say the Read tool reads PDFs longer than 10 pages in ranges such as pages 1-5.
Evidence: Tools referencePublic state history
Every material classification change preserves its prior state, new state, effective research date, and supporting evidence.
Anthropic
The tools reference and the Claude Code changelog both say the Read tool reads PDFs longer than 10 pages in ranges such as pages 1-5.
Evidence: Tools referenceProject Editor docs place Agent in the editor beside a live preview. The agent overview says you start chatting there. Plan names do not match this registry one-for-one.
Evidence: Project EditorCognition
Desktop docs and the Windsurf 2.0 announcement both say delegated cloud Devin keeps working after the laptop is closed. Pro and Teams are the registry plans the docs name.
Evidence: Devin in Devin DesktopAgent launch note documents a terminal on ChatGPT's virtual computer. ChatGPT Work docs document shell commands. The agent help article says agent mode is not on Free.
Evidence: Introducing ChatGPT agentWork with Apps documents IDE file reads and edits from the ChatGPT macOS app. The OpenAI Marketplace listing documents the VS Code connection. No plan list was invented.
Evidence: Work with Apps on macOSAnthropic
Desktop docs document file attachments. The CLI first-day guide says there is nothing to upload or attach. No plan list was named.
Evidence: Desktop applicationAnysphere
Agent changelog says agents can read attached PDFs. Grok Bot docs list PDFs as supported inputs. Plan strings were not copied from the billing page.
Evidence: Subagents, Skills, and Image GenerationAnysphere
Agent changelog documents attaching PDFs in chat. Grok Bot docs list user file inputs. No plan list was invented.
Evidence: Subagents, Skills, and Image GenerationAnthropic
Official docs and the Claude Code power-user guide both describe auto memory and CLAUDE.md as cross-session context. No plan gate is named.
Evidence: How Claude remembers your projectAnysphere
Rules docs establish persistent cross-session instructions. The 1.0 changelog announced per-project Memories as a beta. Current plan scope for automatic Memories is not restated.
Evidence: RulesOfficial memories docs and the Codex pricing matrix establish local memories on the listed plans, region-limited and off by default.
Evidence: MemoriesOfficial Agent docs say image generation is built in with no toggle; the chat asset guide names image generation as a project asset path. Plan tiers are not named.
Evidence: Image GenerationOfficial docs and changelog establish a native Copilot desktop app on every plan for macOS, Windows, and Linux.
Evidence: About the GitHub Copilot appOfficial changelog names Pro and Pro+; docs describe repository and user preference memory across sessions.
Evidence: Copilot Memory now on by default for Pro and Pro+ users in public previewOfficial chat upload docs name asking questions about a PDF; conversations docs cover analysis of uploaded files.
Evidence: Upload and use filesOfficial upload docs name spreadsheet analysis; conversations docs cover uploading spreadsheets to analyze them.
Evidence: Upload and use filesConversation prompt-box file upload documented across independent Replit doc sections.
Evidence: Upload and use filesImage and screenshot attach documented for Agent prompts.
Evidence: Upload and use filesAnysphere
Agent image attach documented on product docs and iOS help.
Evidence: Prompting agentsAnthropic
Image paste and analysis documented for Claude Code product surfaces.
Evidence: Common workflowsCopilot vision GA documents image attach on all plans.
Evidence: Copilot vision is generally availableChat file attach documented for all Copilot plans.
Evidence: Asking GitHub Copilot questions in GitHubAgent Web Search documented as built-in product capability.
Evidence: Web SearchPDF attach listed with Copilot vision GA and Chat docs.
Evidence: Copilot vision is generally availableAnthropic
Official FAQ and vision docs state Claude cannot generate or create images.
Evidence: Can Claude produce images?Official Replit Agent MCP client docs establish product MCP support across Agent-available plans.
Evidence: Connect via MCPAnthropic
Chrome browser control verified for Claude Code on Pro/Max/Team/Enterprise via extension.
Evidence: Use Claude Code with ChromeAnthropic
Built-in WebSearch tool documented for Claude Code product surfaces.
Evidence: Tools referenceBing web search documented across individual and business/enterprise Copilot Chat with policy enablement.
Evidence: Managing GitHub Copilot policies as an individual subscriberAnysphere
Native Browser tool documented product-wide; not limited to external Playwright MCP setup.
Evidence: BrowserAnysphere
Agent image generation documented in product docs and help.
Evidence: Cursor AgentScoped MCP agent mode to Standard/Enterprise and both VS Code and IntelliJ.
Evidence: Gemini Code Assist overviewAnthropic
Dispatch background agents verified for Pro and Max.
Evidence: Run tasks in the background with DispatchOpenAI
Free included for code-backed data analysis with limited quotas.
Evidence: ChatGPT Free Tier FAQCloud agent scoped to paid Copilot plans; Free excluded.
Evidence: Managing access to GitHub Copilot cloud agentAnthropic
Claude in Chrome establishes browser control on paid Claude plans.
Evidence: Get started with Claude in ChromeScheduled Tasks establish asynchronous background work on named paid plans including Edu.
Evidence: Scheduled Tasks in ChatGPTAnthropic
Agent view establishes leave-session background work; support release notes corroborate Claude Code Dispatch while away.
Evidence: Manage multiple agents with agent viewOpenAI
File upload scoped across Free through Enterprise/Edu; Free is limited rather than excluded.
Evidence: ChatGPT Free Tier FAQOpenAI
PDF analysis available across Free through Enterprise/Edu with Free limited quotas.
Evidence: File Uploads FAQSpreadsheet analysis documented across Free (limited) and paid plans via data analysis plus Excel add-in scope.
Evidence: Data analysis with ChatGPTOpenAI
Image input present on Free through paid plans; Free limited.
Evidence: ChatGPT Free Tier FAQThird-party apps/plugins available across Free through Enterprise/Edu with plan- and admin-specific gates.
Evidence: Apps in ChatGPTPerplexity AI
Perplexity documented asynchronous work, persistent context, and scheduled monitoring in the Computer surface.
Evidence: What is Computer?Google documented MCP servers in Gemini Code Assist agent mode while retaining a preview label.
Evidence: Gemini Code Assist agent mode overviewOpenAI
OpenAI added an in-app browser for local web development and described broader browser command as a future direction.
Evidence: Codex for (almost) everythingOpenAI renamed connectors to apps and documented a unified directory with file search, deep research, sync, and action capabilities.
Evidence: Apps in ChatGPT