Claude Code
Claude Code is the reference point for terminal-first coding agents: it works from a plain shell, reaches into every editor and chat surface Anthropic ships, and its extension points (MCP, hooks, skills, subagents) are deep enough that teams build their own workflows on it. The buying decision is really about usage: Pro is enough to try it, but sustained agentic work eats the shared five-hour and weekly limits fast, and the realistic budgets are Max at $100-200 or per-token Console billing. Skip it if you want an autocomplete pane or cannot let an agent run commands in your repo.
Pros
- One engine across terminal, VS Code, JetBrains, desktop, web, mobile, GitHub, GitLab and Slack, sharing CLAUDE.md, settings and MCP servers
- Extension points for real customisation: MCP servers, hooks, skills, plugins, parallel subagents and an Agent SDK
- Auto mode with checks on risky commands, scheduled routines and cloud sessions that keep running when the laptop is closed
- Included in every paid Claude plan from $17/month, with Max tiers bundling API credits
Cons
- No free tier, and the Free plan explicitly excludes it
- Usage comes out of the same rolling five-hour and weekly limits as chat, so heavy sessions hit the ceiling on Pro
- Running it at scale means Max at $100-200 a month or per-token API spend that is hard to forecast
- IDE extensions and the Linux desktop app still carry beta labels