IDE agent Introduced in L5

Cursor

An AI-first IDE — VS Code fork with deep agent integration. Edit-with-AI, tab completions, repo-aware chat.

Mindmap

hover · click to navigate
this tech depends on / used by alternative Shipyard anchor
What it is

The plain-English version

Cursor is an IDE built around AI coding. A fork of VS Code with custom features: in-line chat, multi-file edits, tab-completion driven by an LLM, codebase-aware chat that can search and reason across the whole project. Subscription-based; works with multiple model providers.

Why it exists

The problem it solves

If you live in an editor and want AI woven into every interaction (not a separate chat window), Cursor's UX is hard to beat. The Composer multi-file edit flow, the in-line edit mode, and the tab autocomplete all reduce friction. AI-first IDE has become a category, with Cursor leading.

What it competes with

Alternatives

AlternativeTypeWhen it wins
Claude CodeCLI agentAnthropic's CLI coding agent. Runs in your terminal, in your repo, with full project context.
Where it shows up in Shipyard

Deep links

Vocabulary

The words you'll hear

Composer
Multi-file edit mode. Specify the change; Cursor edits across files.
Tab autocomplete
Aggressive next-line predictions. Press Tab to accept.
@ symbols
@file, @codebase, @docs — pull context into chat.
Cmd-K / Cmd-L
Inline edit / chat toggle. Muscle memory builds fast.
.cursorrules
Project-level instructions, similar to CLAUDE.md.
Apply / Reject
Review each suggested edit before it lands.
Prompting

Bad vs. good prompt for Cursor

✕ Bad prompt
improve this file
✓ Good prompt
Add input validation to the createTask handler. Use zod, define the schema in src/schemas/task.ts, import it in the handler, return 400 with the formatted error if invalid. Don't edit any other handlers in this file. After the edit, run `npm run typecheck` and tell me if it passes.

Why it works: Specifies the library, the file structure, the exact handler to touch, the bound (don't touch others), and the verification step. Cursor's Composer handles this cleanly with these constraints.

Pitfalls

What bites real teams

⚠ Tab-completion trust

Autocomplete fills in plausible-looking code that's slightly wrong. Read what you accept.

⚠ Composer over-edit

Without scope constraints, Composer will rewrite more than you wanted. Always specify what to leave alone.

⚠ No .cursorrules

Without project rules, the AI re-discovers conventions each session. Set them up early.

References

Official docs only