GraphQL

The HTTP client’s Body type → GraphQL mode gives you:

• A Query field (multi-line code editor with bracket matching)
• A Variables field (JSON)

Posts a { query, variables } envelope as application/json.

Schema browser

A left-rail panel runs __schema introspection against the endpoint on change (debounced ~600ms, cached per URL). Types render root-first, then grouped by kind, with one-click expand and a search box. If the server has introspection disabled, the panel says so honestly instead of looping.

Autocomplete

Context-aware autocomplete runs on top of the introspected schema — no external parser, no extra dependency.

• Triggers on identifier keystrokes and {, (, :, $
• Four contexts detected by walking the buffer up to the cursor: top-level operation keywords, selection-set fields, argument lists, and type-name positions in variable declarations
• Kind badges on every row — fields (emerald), args (cyan), types (amber)
• Smart insert — fields with required arguments insert as fieldName(, arguments insert as argName: so you stay in flow
• Keyboard — ↑↓ to navigate, Tab or Enter to accept, Esc to dismiss, Ctrl+Space to force-invoke
• Filter is case-insensitive with prefix matches ranked above substring matches

The scanner tracks balanced braces and parens, line and triple-quoted strings, # comments, alias prefixes (alias: field {) and pre-brace arg lists (field(arg: $v) {), so the field-name stack stays correct in deeply nested queries.

What you get

• Saving the request into a collection
• Running it through the runner
• Environment / {{var}} substitution in the query
• SSH-tunnel routing

What’s separate

The dedicated GraphQL tab on the API rail is a scratchpad — no persistence. Use it to noodle; save into the HTTP client when you want to keep a query.