Skip to main content

Overview

Agents are Claude Code SDK instances running within a workspace’s git worktree. You see every message, tool call, file edit, and cost in real time.

Starting Agents

Start an agent by writing a prompt in a workspace. You can run multiple agents per workspace for parallel tasks. Each message supports per-message overrides:
OptionValuesDescription
ModelDefault, Opus, Opus 1M, Sonnet, HaikuOverride the default model. Opus 1M uses a 1 million token context window.
EffortLow, Medium, High, MaxControl thoroughness
ThinkingOn/OffExtended thinking mode
Fast ModeOn/OffFaster output
Plan ModeOn/OffStep-by-step with approval

Lifecycle Controls

ActionDescription
PauseSuspend the agent, preserving session
ResumeContinue from where it paused
KillTerminate immediately
WarmupPre-warm an agent session for faster response on the next message
RenameChange the agent’s display name
Resume AllResume all paused agents across workspaces
Rerun AllCreate fresh agents for all workspace slots

Message Queue

When an agent is busy, messages are queued:
  • Queued messages are persisted to SQLite (crash-safe)
  • Auto-delivered when the agent becomes available
  • Can be cancelled or force-delivered
  • Each queued message preserves its model/effort/thinking/fast mode options
  • Pending messages appear in a visual queue below the input area

Plan Mode

When enabled, agents propose a step-by-step plan before executing:
  1. Agent generates plan steps
  2. You review and approve (or edit)
  3. Agent executes approved steps
  4. Progress tracked per-step

Session Resumption

Agent sessions persist across app restarts:
  • Session IDs stored in state
  • On restart, agents with valid sessions resume automatically
  • Agents without sessions are marked as errored with “Interrupted by app restart”
  • Restart count is tracked per agent

Token & Cost Tracking

Per-agent tracking includes:
  • Input/output token counts
  • Cache creation and read tokens
  • Cost in USD (regular and overage costs tracked separately)
  • Context window fill percentage with visual indicator
  • Daily aggregate usage across all agents
Configure Always Show Context Usage in Settings to display the context fill percentage at all times. By default, it only appears when the context window is more than 70% full.

Usage & Rate Limits

The status bar shows your Claude account usage and rate limit status:
  • Session (5hr) — Rolling 5-hour token usage
  • Weekly — 7-day aggregate usage
  • Weekly Opus / Weekly Sonnet — Per-model weekly limits
  • Overage — Usage beyond your subscription limits
Each bucket shows a utilization bar and resets-at time. Click the usage indicator in the status bar to see the full breakdown.

Screenshot Sharing

Share visual context with agents:
  • Capture screenshots using the global hotkey
  • Drag and drop images into the message input
  • Recent screenshots appear in a strip below the input
  • Agents can reference screenshots when making decisions

Question Cards

Agents and the manager can ask you structured questions:
  • Single choice — Pick one option from a list
  • Multiple choice — Select multiple options
  • Free text — Type a response
Questions appear as interactive cards in the conversation. Answer directly or dismiss to skip.

MCP Server Status

View the connection status of MCP servers configured for your repository:
  • Connected, disconnected, or error states shown per server
  • Click to view details or troubleshoot