smart_read

Use smart_read when you know the exact 1-based inclusive ranges to inspect.

Use when

A diagnostic or review identifies precise lines.
You need a small set of ranges from one file.

Do not use when

You need to discover containing symbols; use line_context or outline.
You need a definition from a visible identifier; use jump.

Request

{"file":"/workspace/project/src/lib.rs","lines":[{"start":40,"end":58}],"output_mode":"type_annotated","workspace_root":"/workspace/project"}

Parameters

Field	Required	Default / bounds
`file`	yes	Path (`file_path` is accepted as an alias).
`lines`	no	`[]` reads the complete file (reserve for small files); each `start` ≥ 1.
`lines[].end`	no	Omitted means through end of file; ranges are 1-based and inclusive.
`output_mode`	no	`type_annotated`; union: `type_annotated`, `plain_text`, `both`.
`workspace_root`	no	Only needed when multiple workspaces are active.

What it returns

The requested file ranges. Type-annotated output uses resolved LSP inlay hints and falls back to plain text with a short per-range note when hints are unavailable.

Example

Real output — a plain ranged read, still annotated:

53|impl UpstreamError {
54|    pub(crate) fn into_mcp_error(self) -> McpError {
55|        match self {
56|            Self::Mcp(error/*: ErrorData*/) => error,
57|            Self::Cancelled => McpError::internal_error(/*message: */"daemon request cancelled", /*data: */None),

Resolved types and call-site parameter names, even on the simplest read there is.

Common mistakes

Using zero-based or exclusive line numbers.
Omitting end unintentionally on a large file.
Reading an entire file when outline plus extract would be more targeted.

line_context finds the containing symbol when only a line is known.
outline maps a file before deciding what to read.
extract reads by symbol identity instead of line coordinates.

On this page