Use when working with TypeScript or JavaScript code and typescript-lsp plugin is enabled - for finding references, checking types, navigating definitions, or verifying type correctness
Using TypeScript LSP
Overview
The typescript-lsp plugin provides semantic code intelligence for TypeScript and JavaScript. Use LSP tools instead of grep/read when you need semantic accuracy - LSP understands TypeScript's type system, inheritance, interfaces, and symbol relationships.
When to Use LSP vs Text Search
digraph lsp_decision {
"Need code info?" [shape=diamond];
"Semantic accuracy needed?" [shape=diamond];
"Use LSP tool" [shape=box];
"Use Grep/Read" [shape=box];
"Need code info?" -> "Semantic accuracy needed?" [label="yes"];
"Need code info?" -> "Use Grep/Read" [label="no - text patterns"];
"Semantic accuracy needed?" -> "Use LSP tool" [label="yes"];
"Semantic accuracy needed?" -> "Use Grep/Read" [label="no"];
}
Use LSP for:
- Finding all usages before refactoring
- Getting type information (inferred or explicit)
- Navigating to definitions (including node_modules)
- Checking for type errors
- Understanding interface implementations
Use Grep for:
- Text patterns (comments, strings, config files)
- Cross-language searches
- JSON, YAML, markdown content
- When LSP unavailable
LSP Operations Quick Reference
| Task | LSP Operation | Instead of |
|---|---|---|
| Find all callers of a function | findReferences |
Grep for function name |
| Get function/variable type | hover |
Read file + parse |
| Jump to definition | goToDefinition |
Grep + Read |
| Check type errors | getDiagnostics |
Run tsc CLI |
| View file structure | documentSymbol |
Read entire file |
Key Patterns
Before Refactoring: Find All References
Wrong approach:
Grep for "formatDuration" -> may miss renamed imports, get false positives from comments
Right approach:
LSP findReferences on symbol -> gets ALL semantic usages, no false positives
Getting Type Information
Wrong approach:
Read file -> find function -> manually trace type inference
Right approach:
LSP hover on symbol -> instant type signature with inferred types resolved
Verifying Type Correctness
Wrong approach:
Bash: npx tsc --noEmit (slow, external process)
Right approach:
LSP getDiagnostics on changed files -> instant, no subprocess
Finding Interface Implementations
Wrong approach:
Grep for "implements InterfaceName" -> misses duck-typed objects
Right approach:
LSP findReferences on interface -> finds all implementations including structural matches
Common Mistakes
| Mistake | Why It's Wrong | Fix |
|---|---|---|
| Grep for symbol before rename | Text search misses aliased imports, re-exports | Use findReferences |
| Read entire file for one type | Wastes context, slow | Use hover |
| Run tsc after every change | Slow, external tool | Use getDiagnostics |
| Grep to find definition | May find wrong match (same name, different module) | Use goToDefinition |
| Grep for interface usages | Misses structural/duck typing | Use findReferences |
How LSP Tools Work
LSP is a built-in Claude Code tool, not a bash command. When typescript-lsp plugin is enabled, Claude Code automatically provides these operations:
goToDefinition- navigate to where symbol is definedfindReferences- find all usages of a symbolhover- get type signature and documentationgetDiagnostics- get type errors for a filedocumentSymbol- list all symbols in a file
You invoke these through Claude's tool system, not via bash. If LSP is unavailable, fall back to grep/read but document the limitation.
Supported File Types
.ts, .tsx, .js, .jsx, .mts, .cts, .mjs, .cjs
Prerequisites
- Plugin enabled:
typescript-lsp@claude-plugins-officialin~/.claude/settings.json - Language server installed:
npm install -g typescript-language-server typescript - Project configured:
tsconfig.jsonorjsconfig.jsonpresent (recommended for accuracy)
Checking LSP Availability
Always try LSP first. If the LSP tools aren't available, they'll return an error. Don't preemptively fall back to grep just because you're unsure.
1. Attempt LSP operation (e.g., findReferences)
2. If tool error → fall back to grep/read with documented limitation
3. If success → use the semantic results
Fallback Strategy
If LSP is unavailable (tool not found, server not running):
- Document the limitation - explicitly state: "LSP unavailable, using text search as fallback"
- Use Grep with caution - understand it may miss aliased imports or re-exports
- Verify manually - for refactoring, double-check results make sense semantically
- Suggest fix - remind user to ensure typescript-language-server is installed and plugin enabled
Never silently fall back. Always tell the user when you're using text search instead of LSP so they understand the limitations.
Red Flags - Stop and Use LSP
These thoughts mean you should use LSP, not grep:
| Thought | Reality |
|---|---|
| "Let me grep for this function" | If it's a symbol, use findReferences |
| "I'll read the file to find the type" | Use hover for instant type info |
| "Let me run tsc to check" | Use getDiagnostics - faster, no subprocess |
| "I'll search for where this is defined" | Use goToDefinition - semantic accuracy |
| "Grep is faster for quick searches" | LSP is faster for semantic queries |
| "I need to find all implementations" | Use findReferences - catches structural matches |
When LSP Won't Help
- No typescript-language-server binary installed
- Working with pure JavaScript without jsconfig.json (limited type inference)
- Cross-language references (e.g., TypeScript calling Rust via WASM)
- Comment/string literal searches
- Config files (JSON, YAML, etc.)
- Node modules not installed (
npm installneeded first)
You Might Also Like
Related Skills

coding-agent
Run Codex CLI, Claude Code, OpenCode, or Pi Coding Agent via background process for programmatic control.
openclaw
add-uint-support
Add unsigned integer (uint) type support to PyTorch operators by updating AT_DISPATCH macros. Use when adding support for uint16, uint32, uint64 types to operators, kernels, or when user mentions enabling unsigned types, barebones unsigned types, or uint support.
pytorch
at-dispatch-v2
Convert PyTorch AT_DISPATCH macros to AT_DISPATCH_V2 format in ATen C++ code. Use when porting AT_DISPATCH_ALL_TYPES_AND*, AT_DISPATCH_FLOATING_TYPES*, or other dispatch macros to the new v2 API. For ATen kernel files, CUDA kernels, and native operator implementations.
pytorch
skill-writer
Guide users through creating Agent Skills for Claude Code. Use when the user wants to create, write, author, or design a new Skill, or needs help with SKILL.md files, frontmatter, or skill structure.
pytorch
implementing-jsc-classes-cpp
Implements JavaScript classes in C++ using JavaScriptCore. Use when creating new JS classes with C++ bindings, prototypes, or constructors.
oven-sh
implementing-jsc-classes-zig
Creates JavaScript classes using Bun's Zig bindings generator (.classes.ts). Use when implementing new JS APIs in Zig with JSC integration.
oven-sh