
Install Claude Code, run your first real task, and connect tools via MCP, plus the honest cost and the zero-config alternative if setup is the blocker.

A command-line agent. You type what you want in plain English inside a project directory, and Claude Code reads the codebase, proposes changes, edits files and runs commands, asking your permission as it goes. It also exists in VS Code and as a web version, but the terminal is the canonical experience.
You need Node.js 18+, then:
npm install -g @anthropic-ai/claude-code
Or use the native installer from Anthropic's docs if you prefer not to use Node. You also need a Claude subscription (Pro, Max, Team or Enterprise) or an Anthropic API key with billing enabled; Claude Code does not work on the free tier.
cd your-project
claude
It opens an interactive session. First launch asks you to authenticate and trust the folder. From there, just describe what you want: "find where the login redirect is handled and explain it" or "add a test for the retry logic in utils.ts".
/init — makes Claude write a CLAUDE.md with your project's conventions; do this once per repo./permissions — control what it may run without asking (file edits, specific bash commands)./clear — reset context when switching tasks; long sessions get slower and pricier./cost — check token spend so far./agents — manage subagents for parallel work.Esc — interrupt it mid-action.Claude Code connects to outside apps through MCP servers. The command is:
claude mcp add <name> <command-or-url>
Examples that work today:
claude mcp add --transport http linear https://mcp.linear.app/mcp — Linear's hosted server, OAuth sign-in.claude mcp add --transport http notion https://mcp.notion.com/mcp — Notion's hosted server.claude mcp add github -- npx -y @github/github-mcp-server — needs a GitHub token in the env.claude mcp add slack -- npx -y @modelcontextprotocol/server-slack — needs a Slack bot token you create yourself.http://127.0.0.1:3845/mcp (requires a paid Dev/Full seat).Each tool means another server, another credential, another config block. It is powerful, but it is also why most people stop at one or two connections.
cd into a repo you know./init and skim the CLAUDE.md it writes.On a Claude Pro ($20/mo) plan you share limits with the Claude app; heavy coding sessions hit them fast. Max ($100+/mo) or API billing is what sustained daily use realistically costs. This is the honest catch: Claude Code is excellent and also the most expensive way to talk to Claude.
If what you actually want is "an AI agent that uses my tools" without a terminal, Node install, MCP configs or token juggling: Krater does the same pattern in the browser. The agent already has Gmail, Slack, Notion, GitHub, Linear, Figma, Drive, Calendar and 30+ more apps wired in, you click once to connect each, and it picks the right tool from your request. Different surface, same idea, no config files.
No. It requires a paid Claude plan (Pro, Max, Team or Enterprise) or API billing. There is no free tier for it.
The app for questions, writing and general chat. Claude Code when the work lives in files and repos: features, refactors, debugging, anything where "run the tests and show me" is the answer.
Not directly. Web app connectors exist for the Claude app, but in Claude Code, MCP servers are the tool path. The alternative is an agent that ships with the integrations already connected.
No, it is Anthropic's own tool. If you want one agent that can switch between Claude, GPT, Gemini and 500+ other models while still using your connected apps, that is the Krater pitch exactly.