Claude: the assistant I build with.

What is Claude AI?

Claude is a family of large language models and the assistant built on them, developed by Anthropic. You can chat with it on the web, on desktop and mobile apps, and inside tools such as Slack and code editors; developers call the same models through an API. The models come in tiers by capability and price, with reasoning modes, large context windows, image input and tool use. Anthropic emphasises safety research and careful behaviour in how it trains Claude, which shows in its tendency to explain its reasoning, acknowledge uncertainty and decline clearly rather than guess.

Who makes Claude AI?

Anthropic, an AI safety and research company founded in 2021 by former OpenAI researchers, including Dario and Daniela Amodei, and headquartered in San Francisco. It trains the Claude models, runs the Claude apps and API, publishes research on model behaviour and interpretability, and created the Model Context Protocol, the open standard for connecting AI assistants to tools that the wider industry adopted. The company's models are also offered through cloud platforms such as Amazon Bedrock and Google Cloud's Vertex AI for customers who want them inside those ecosystems.

How long has Claude AI been around?

The first Claude model was released in March 2023, with Claude 2 later that year and the Claude 3 family in early 2024, followed by successive generations that improved reasoning, coding and context handling. The agentic coding tool and the Model Context Protocol arrived in 2024 and 2025. So Claude has existed for a little over three years as of 2026, which in this field is a long time; the capability gap between the first version and the current one is enormous. The constant across versions has been the emphasis on careful, honest behaviour.

What is Claude API?

The Claude API is Anthropic's developer interface: send messages (text, images, documents) with a system prompt and options, receive the model's reply, streamed or whole, including tool calls and structured output. It supports prompt caching to cut cost on repeated context, batch processing at a discount, extended thinking for hard problems, citations over provided documents, and files. Official libraries exist for Python and TypeScript, with others from the community, and an agent SDK packages the loop used by Anthropic's own coding tool. The same API is available through Bedrock and Vertex AI with those platforms' billing.

How to use Claude API

• Create an account on the developer console, add credits, and generate an API key.
• Store the key in an environment variable or an ignored env file; never in code.
• Install the client library in a virtual environment.
• Send a messages request: a system prompt, a user message, the model name and a maximum token count; print the reply.
• Add streaming for chat interfaces, tool definitions for actions, and cache control on long stable prefixes to reduce cost.
• Handle rate limits with retries; log token usage per request; set a spending limit in the console.

What does Claude AI do?

Claude reads and writes: it answers questions, summarises and analyses long documents, drafts and edits text, writes and debugs code, reasons through problems step by step, describes images, and, through tools and the Model Context Protocol, acts on external systems such as files, databases and web services. In the chat app it works as a conversational assistant with projects and file uploads; through the API it becomes a component inside your own software; through its agentic coding tool it edits codebases and runs commands under your supervision. It does not browse the web on its own unless given a search tool, and it does not generate images.

Can I use Claude AI for free?

Yes, the chat app has a free plan: create an account and you get daily usage with the default model, file uploads and projects, with limits that reset over time. Paid plans raise the limits, unlock the strongest models and reasoning modes, and include the coding tool. The API has no free tier; it runs on credits. For free API-style experimentation, use open models locally with Ollama or a provider with a free tier, then switch to Claude when you need its quality for a real task; the client code changes little if you keep provider calls behind one interface.

How much Claude AI cost

Two separate prices. The chat app: a free plan, and paid individual and team plans billed monthly, with the current amounts on Anthropic's pricing page. The API: billed per million input and output tokens, with each model tier at a different rate, cached input tokens heavily discounted, and batch requests discounted further; the numbers change with each model generation, so check the page. To estimate, count a typical request's tokens and multiply; prompt caching on a long system prompt or document set often cuts the input bill substantially, which is why the context and caching pages on this site matter for cost.

When to use Claude AI

• Reading and reasoning over long documents, contracts, codebases and transcripts.
• Writing that needs structure, restraint and accuracy: documentation, reports, explanations.
• Coding, especially multi-file changes through the agentic coding tool.
• Agents and tools: the API's tool use, the agent SDK and MCP make it a natural base for agentic systems.
• Tasks where honest uncertainty matters more than confident fluency.
• Less so for: image generation (it does not make images) or the widest consumer integrations, where other assistants lead.

Can Claude AI build a website?

Yes, within limits that depend on how you use it. In the chat app it can write a complete single-page site and show it as a live preview. With the agentic coding tool it can scaffold a full project, write components, run the build, fix errors and deploy, working across files like a developer. This site is an example: much of its code, its MCP server and the pipeline that published these pages were built in collaboration with Claude through that tool. It still needs a person who knows what they want, reads the diffs, and tests; the model proposes, you decide. For a student, it is the fastest way to learn web development by building something real.