Skip to main content

​
Overview

The hive tui command launches an interactive terminal user interface (TUI) for monitoring and controlling agent execution. 
hive tui [options]

​
Options

​
--model
str
LLM model to use (any LiteLLM-compatible name)
hive tui --model claude-sonnet-4-20250514

​
Features

​
Agent Picker

When you launch the TUI without specifying an agent, you get an interactive agent picker:
  • Browse available agents in exports/
  • See agent descriptions and metadata
  • Select an agent to run

​
Live Execution Monitor

Once an agent is loaded: Execution View:
  • Real-time node execution status
  • Current node being processed
  • Execution path visualization
  • Step count and timing
Conversation Panel:
  • Live LLM conversations
  • Tool calls and results
  • Decision reasoning
  • Input/output for each node
Status Bar:
  • Current status (running, paused, completed, failed)
  • Token usage
  • Execution time
  • Model being used

​
Interactive Input

The TUI includes a chat-style interface:
  • Enter input naturally (auto-formatted with Haiku)
  • Support for JSON input
  • Session memory across inputs
  • Special commands (see below)

​
Session Management

  • Resume from paused sessions
  • Resume from checkpoints
  • View session history
  • Export session data

​
Keyboard Shortcuts

​
Tab
key
Cycle between panels (chat input, conversation history, node view)
​
Ctrl+C
key
Graceful shutdown (saves session state)
​
Ctrl+D
key
Exit TUI
​
Ctrl+P
key
Pause execution (at next node transition)
​
Ctrl+R
key
Resume paused execution
​
Ctrl+L
key
Clear screen
​
Up/Down
key
Navigate conversation history
​
PgUp/PgDn
key
Scroll conversation panel

​
Special Commands

While in the TUI, you can enter special commands in the input field:
​
/help
command
Show help and available commands
​
/info
command
Show agent information (name, goal, nodes, edges)
​
/nodes
command
List all nodes in the agent graph
​
/status
command
Show current execution status
​
/pause
command
Pause execution at next node transition
​
/resume
command
Resume paused execution
​
/reset
command
Reset conversation state and agent session
​
/sessions
command
List available sessions for this agent
​
/checkpoints
command
List checkpoints for current session
​
/export
command
Export current session data to file
​
/quit
command
Exit TUI (same as Ctrl+D)

​
Examples

​
Launch TUI with Agent Picker

hive tui
Screen: 
┌───────────────────────────────────────────────┐
│ Hive Agent Dashboard                           │
├───────────────────────────────────────────────┤
│ Available Agents:                              │
│                                                │
│ > Calculator                                   │
│   Perform mathematical calculations            │
│                                                │
│   Lead Qualifier                               │
│   Qualify sales leads                          │
│                                                │
│   Email Sender                                 │
│   Send personalized emails                     │
│                                                │
└───────────────────────────────────────────────┘

Use ↑↓ to select, Enter to load, Ctrl+C to exit

​
Launch with Specific Agent

hive run exports/calculator --tui
Screen (during execution): 
┌────────────────────────────────────────────────────────────────────────────────┐
│ Calculator - Running                                                    │
├────────────────────────────────────────────────────────────────────────────────┤
│ Execution Path:                                                         │
│ ✓ input_parser → ● calculator → output_formatter                    │
│                                                                          │
├────────────────────────────────────────────────────────────────────────────────┤
│ Conversation:                                                            │
│                                                                          │
│ [input_parser]                                                           │
│ Parsing expression: "2 + 3"                                              │
│ Tool call: parse_expression(expression="2 + 3")                          │
│ Result: {"op": "+", "left": 2, "right": 3}                              │
│                                                                          │
│ [calculator]                                                             │
│ Calculating: 2 + 3                                                       │
│ Tool call: calculate(op="+", left=2, right=3)                            │
│ Result: 5                                                                │
│                                                                          │
├────────────────────────────────────────────────────────────────────────────────┤
│ >>> Enter input (or /help for commands)                                  │
└────────────────────────────────────────────────────────────────────────────────┘

Status: Running | Steps: 2/100 | Tokens: 150 | Time: 1.2s | Model: claude-haiku-4-5

​
Resume from Session

hive resume exports/assistant session_123 --tui
The TUI loads the session state and shows the conversation history before accepting new input.

​
Using Special Commands

hive tui
# Select agent
>>> /info
Output: 
┌────────────────────────────────────────────────────────────────────────────────┐
│ Agent Information                                                       │
├────────────────────────────────────────────────────────────────────────────────┤
│ Name: Calculator                                                         │
│ Goal: Perform mathematical calculations accurately                       │
│ Nodes: 3                                                                 │
│ Edges: 2                                                                 │
│                                                                          │
│ Success Criteria:                                                        │
│   - Result matches expected mathematical answer                          │
│                                                                          │
│ Constraints:                                                             │
│   - [HARD] Handle invalid inputs gracefully                              │
└────────────────────────────────────────────────────────────────────────────────┘

​
Panels

The TUI is organized into several panels:
  • Agent name
  • Current status
  • Model being used

​
Execution Path

  • Visual representation of node flow
  • Current node highlighted
  • Completed nodes marked with ✓

​
Conversation History

  • Scrollable conversation log
  • LLM messages
  • Tool calls and results
  • Decision reasoning
  • Timestamps

​
Input Field

  • Chat-style input
  • Command completion
  • Input history (Up/Down arrows)

​
Status Bar

  • Execution state
  • Step count
  • Token usage
  • Elapsed time
  • Model name

​
Color Scheme

  • Green: Success states, completed nodes
  • Yellow: Running, in-progress
  • Red: Errors, failed nodes
  • Blue: Information, node names
  • Cyan: Tool calls, special commands
  • Gray: Timestamps, metadata

​
Exit

To exit the TUI:
  1. Ctrl+D or Ctrl+C: Graceful shutdown (saves session state)
  2. /quit command: Exit via command
The TUI automatically saves the current session state before exiting, allowing you to resume later.

​
Advanced Features

​
Multi-Panel Layout

Press Tab to cycle focus between:
  1. Input field
  2. Conversation history (scrollable)
  3. Node view (if available)

​
Session Export

Export the current session data: 
>>> /export session_data.json
Exports:
  • Conversation history
  • Execution path
  • Memory state
  • Metadata

​
Real-Time Updates

The TUI updates in real-time:
  • Node transitions appear instantly
  • Token usage updates live
  • Streaming output shows as it’s generated

​
Troubleshooting

TUI doesn’t render correctly:
  • Ensure terminal supports 256 colors
  • Minimum terminal size: 80x24
  • Try resizing terminal window
Input lag:
  • Disable verbose logging
  • Reduce conversation history size
  • Use --quiet mode
Can’t see full output:
  • Use PgUp/PgDn to scroll
  • Increase terminal size
  • Export session and view in editor

Build docs developers (and LLMs) love

Get started for freeTalk to us