CLI Status Line

The CLI supports a user-configurable status line rendered above the prompt. A command is spawned on each conversation update, receives a JSON payload on stdin describing the session, and its stdout is displayed as the status line. The spec is aligned with Claude Code's status line.

Configuration

Add a statusLine entry to ~/.cursor/cli-config.json:

{
  "statusLine": {
    "type": "command",
    "command": "~/.cursor/statusline.sh",
    "padding": 2
  }
}

The command field supports full paths, ~ expansion, and shell-style argument splitting. You can point it at a script file or use an inline command like jq -r '...'.

Field	Required	Default	Description
type	yes	—	Must be "command"
command	yes	—	Path to an executable or inline command. ~ is expanded.
padding	no	0	Horizontal inset (in characters) for the status line container.
updateIntervalMs	no	300	Minimum interval between invocations. Clamped to >= 300ms.
timeoutMs	no	2000	Maximum time the command may run before it is killed.

Stdin payload

The command receives a JSON object on stdin. The TypeScript interface is StatusLinePayload in packages/agent-cli/src/hooks/use-status-line.ts.

Full JSON schema

{
  "session_id": "abc123",
  "session_name": "my session",
  "transcript_path": "/path/to/transcript.jsonl",
  "render_width_chars": 120,
  "cwd": "/Users/me/project",
  "autorun": false,
  "model": {
    "id": "claude-4-opus",
    "display_name": "Claude 4 Opus",
    "param_summary": "(Thinking)",
    "max_mode": true
  },
  "workspace": {
    "current_dir": "/Users/me/project",
    "project_dir": "/Users/me/project/.cursor/transcripts",
    "added_dirs": []
  },
  "version": "1.2.3",
  "output_style": {
    "name": "default"
  },
  "context_window": {
    "total_input_tokens": 15234,
    "total_output_tokens": null,
    "context_window_size": 200000,
    "used_percentage": 34.5,
    "remaining_percentage": 65.5,
    "current_usage": null
  },
  "vim": {
    "mode": "NORMAL"
  },
  "worktree": {
    "name": "my-feature",
    "path": "/Users/me/.cursor/worktrees/repo/my-feature"
  }
}

Available fields

Field	Description
session_id	Unique session identifier
session_name	Custom session name. Absent if no name has been set
transcript_path	Path to conversation transcript file
render_width_chars	Usable terminal columns minus built-in padding
cwd, workspace.current_dir	Current working directory (both contain the same value)
autorun	true when auto-run is enabled for the current session
workspace.project_dir	Directory where transcripts are stored
workspace.added_dirs	Additional directories (empty array for now)
model.id, model.display_name	Current model identifier and display name
model.param_summary	Formatted parameter summary (e.g. "(Thinking)", "High"). Absent when empty
model.max_mode	true when max mode is enabled. Absent otherwise
version	CLI version string
output_style.name	"default" or "compact"
context_window.total_input_tokens	Estimated input tokens (derived from used_percentage)
context_window.total_output_tokens	Cumulative output tokens (null when not tracked)
context_window.context_window_size	Maximum context window size in tokens
context_window.used_percentage	Percentage of context window used
context_window.remaining_percentage	Percentage of context window remaining
context_window.current_usage	Token counts from the last API call (null before first call)
vim.mode	"NORMAL" or "INSERT" when vim mode is enabled
worktree.name	Worktree name when running inside a worktree
worktree.path	Absolute path to the worktree directory

Fields that may be absent

• session_name — only present when a custom name has been set
• model.param_summary — only present when model has non-default parameters
• model.max_mode — only present when max mode is enabled
• vim — only present when vim mode is enabled
• worktree — only present when running in a worktree

Fields that may be null

• context_window.current_usage — null before the first API call
• context_window.used_percentage, context_window.remaining_percentage — may be null early in the session

Stdout / rendering

• Multiple lines are supported: each line of stdout renders as a separate row in the status area.
• ANSI color codes are supported (use chalk, tput, \033[32m, etc.).
• If the command exits non-zero with empty stdout, the status line is not updated (previous text is kept).
• If the command times out or a new update arrives while the script is running, the in-flight process is killed.
• The status line runs locally and does not consume API tokens.

Examples

Basic: model + context usage

#!/usr/bin/env bash
payload=$(cat)
model=$(echo "$payload" | jq -r '.model.display_name')
pct=$(echo "$payload" | jq -r '.context_window.used_percentage // 0' | cut -d. -f1)
printf "\033[90m%s  ctx %s%%\033[0m" "$model" "$pct"

Context progress bar

#!/usr/bin/env bash
input=$(cat)
MODEL=$(echo "$input" | jq -r '.model.display_name')
PCT=$(echo "$input" | jq -r '.context_window.used_percentage // 0' | cut -d. -f1)

BAR_WIDTH=10
FILLED=$((PCT * BAR_WIDTH / 100))
EMPTY=$((BAR_WIDTH - FILLED))
BAR=""
[ "$FILLED" -gt 0 ] && printf -v FILL "%${FILLED}s" && BAR="${FILL// /▓}"
[ "$EMPTY" -gt 0 ] && printf -v PAD "%${EMPTY}s" && BAR="${BAR}${PAD// /░}"

echo "[$MODEL] $BAR $PCT%"

Multi-line with git info

#!/usr/bin/env bash
input=$(cat)
MODEL=$(echo "$input" | jq -r '.model.display_name')
DIR=$(echo "$input" | jq -r '.workspace.current_dir')
PCT=$(echo "$input" | jq -r '.context_window.used_percentage // 0' | cut -d. -f1)

BRANCH=""
git rev-parse --git-dir > /dev/null 2>&1 && BRANCH=" | 🌿 $(git branch --show-current 2>/dev/null)"

echo -e "\033[36m[$MODEL]\033[0m 📁 ${DIR##*/}$BRANCH"
echo -e "ctx $PCT%"

Inline jq command (no script file)

{
  "statusLine": {
    "type": "command",
    "command": "jq -r '\"[\\(.model.display_name)] \\(.context_window.used_percentage // 0)% context\"'"
  }
}

Testing

Test a script with mock input:

echo '{"model":{"display_name":"Opus"},"context_window":{"used_percentage":25}}' | ./statusline.sh

The command is spawned with child_process.spawn (no shell on Unix, shell: true on Windows for .cmd/.bat compatibility). Updates are debounced at the configured interval. If a new update triggers while a script is running, the in-flight process is killed via AbortController and the new invocation starts immediately.