The interface letting developers directly call Claude capabilities through code, integrating Claude into any application, automation process, or workflow. Biggest difference from claude.ai web: the API lets you precisely control every parameter (model, <a href="/en/glossary/core-concepts/temperature/">Temperature</a>, <a href="/en/glossary/prompt-techniques/system-prompt/">System Prompt</a>, max output length), embedding Claude into your own products and systems rather than using <a href="/en/glossary/core-concepts/anthropic/">Anthropic</a>'s interface. Get an API Key at console.anthropic.com; billed per token.

The authentication credential that allows your application to call the <a href="/en/glossary/claude-tools/claude-api/">Claude API</a>, formatted as a long string starting with `sk-ant-`. Each API Key is linked to an <a href="/en/glossary/core-concepts/anthropic/">Anthropic</a> account; the system uses it to identify who is making requests, calculate billing, and apply corresponding usage limits. Must be kept secure — leaking it allows others to call the API at your expense.

Claude's feature for displaying interactive content in a separate panel alongside conversation — code can be directly executed, webpages previewed in real time, documents edited separately. Upgrades Claude's output from "text replies" to "ready-to-use deliverables." Appears automatically in the claude.ai interface; no configuration needed.

<a href="/en/glossary/core-concepts/anthropic/">Anthropic</a>'s command-line AI coding assistant, letting Claude directly read code, execute commands, and modify files in your terminal — no copy-pasting, no window switching. The biggest difference from claude.ai web: Claude Code can directly access your local file system and work in a real codebase.

An experimental <a href="/en/glossary/core-concepts/anthropic/">Anthropic</a> feature that lets Claude "look at a screen and operate a computer" like a human — moving the mouse, clicking buttons, typing text, taking screenshots to observe results. Without requiring API integration, Claude can directly operate any application with a graphical interface. A significant breakthrough in <a href="/en/glossary/core-concepts/ai-agent/">AI Agent</a> capability.

<a href="/en/glossary/core-concepts/anthropic/">Anthropic</a>'s desktop AI work assistant built for knowledge workers who don't code. Unlike a normal chat that only replies with text, Cowork actually does the work: it browses the web, edits spreadsheets, builds slide decks, and works on designs, completing a whole multi-step office task end to end. Think of it as the office-work counterpart to <a href="/en/glossary/claude-tools/claude-code/">Claude Code</a>, which is the command-line tool aimed at developers.

Claude Desktop is <a href="/en/glossary/core-concepts/anthropic/">Anthropic</a>'s official desktop application that runs Claude locally on your computer (Windows, macOS) without needing the web version. Its signature feature is <a href="/en/glossary/mcp-tools/mcp-protocol/">MCP (Model Context Protocol)</a> support, allowing you to connect your local tools, files, databases, and give Claude access to real-time data and custom integrations. Compared to the web version, Desktop keeps all conversations locally for better privacy, supports file uploads and drag-and-drop for a smoother workflow, and most importantly enables MCP ecosystems — configure Google Drive, GitHub, Slack connectors once and use them across all Desktop conversations, which the web version cannot do.

<a href="/en/glossary/core-concepts/anthropic/">Anthropic</a>'s enterprise-tier Claude subscription for large organizations, including higher usage quotas, centralized management with SSO integration, guarantee that data won't be used for training, and priority support SLA. Suited for organizations needing large-scale AI deployment with compliance and data security requirements.

A Claude.ai workspace feature for creating conversation spaces with fixed Instructions (system prompts) and uploaded documents. Every new conversation opened within a Project automatically carries these settings without needing to re-establish context. Particularly suited for workflows with repetitive tasks and fixed context requirements.

<a href="/en/glossary/core-concepts/anthropic/">Anthropic</a>'s asynchronous <a href="https://claudecowork-me.com/en/glossary/workflow-automation/batch-processing/" target="_blank">Batch Processing</a> API, priced at 50% of standard real-time API, with the trade-off of no guaranteed immediate response (typically completing within 24 hours). Suited for high-volume tasks not requiring real-time feedback: nightly data analysis, bulk document summarization, offline content generation. Combined with <a href="/en/glossary/core-concepts/prompt-caching/">Prompt Caching</a>, can compress actual costs to 5-15% of list price — the highest-savings <a href="/en/glossary/claude-tools/claude-api/">Claude API</a> usage mode.