Skip to main content
For a visual interface, see Getting started with the Manager.
The Idun CLI provides an interactive Terminal User Interface (TUI) for configuring and running agents, and a set of commands for direct operation. It is included in the idun-agent-engine package.

Installation

pip install idun-agent-engine

Operating modes

Launch the interactive setup with:
idun init
The TUI walks you through agent configuration step by step:
  1. Agent information: Name, framework (LangGraph, ADK), port, Python file, and variable selection
  2. Memory: Checkpoint type (in-memory, SQLite, PostgreSQL)
  3. Observability: Provider selection (Langfuse, Phoenix, LangSmith, GCP)
  4. Guardrails: Add safety filters (bias check, PII detection, ban list, and more)
  5. MCP servers: Configure external tool servers
  6. Validate and run: Preview the YAML config, save, and start the agent
  7. Chat: Test the agent with a built-in chat interface
The TUI saves configuration to ~/.idun/<agent-name>.yaml.

Key commands

CommandDescription
idun initLaunch the interactive TUI for agent configuration
idun agent serveStart an agent from a config source
idun agent serve --source file --path <path>Start from a YAML config file
idun agent serve --source managerStart from the Manager API (requires IDUN_AGENT_API_KEY and IDUN_MANAGER_HOST)

Configuration file structure

The TUI generates a YAML config file with this structure:
~/.idun/my-agent.yaml
server:
  api:
    port: 8008

agent:
  type: LANGGRAPH
  config:
    name: my-agent
    graph_definition: ./agent.py:app
    checkpointer:
      type: memory

observability:
  provider: OFF

guardrails: []

mcp_servers: []

TUI features

Agent variable selection

The CLI parses your Python file and lists all available variables. Select the variable that contains your agent or graph instance.

Configuration preview

Before running, the TUI displays your complete YAML configuration with syntax highlighting and line numbers.

Built-in chat

After starting the agent, the Chat section provides:
  • Real-time messaging with the agent
  • Connection status indicators
  • Thinking indicator while the agent processes
  • Word wrapping for long responses

Server management

The Validate and Run section provides two options:
  • Save and Run: Saves config and starts the agent server. The button changes to Kill Server while running
  • Save and Exit: Saves config to disk without starting. The server must be stopped first

API access

Once deployed, your agent exposes a REST API:
# Health check
curl http://localhost:8008/health

# Invoke agent
curl -X POST http://localhost:8008/agent/invoke \
  -H "Content-Type: application/json" \
  -d '{"session_id": "test-session", "query": "Hello"}'

# API documentation
open http://localhost:8008/docs

Keyboard shortcuts

KeyAction
EnterSend message or select option
TabNavigate between sections and inputs
Arrow keysNavigate lists and options
Ctrl+QExit the CLI (stop the server first)
Last modified on March 22, 2026