Agent Node

Agent Node is the working unit in Agent Network -- it receives tasks, invokes an AI model to process them, and reports results.

Not sure which Runtime to pick?

• Want AI to write code / run commands --> codex-sdk (recommended for beginners)
• Want AI to write copy / translate / analyze (programmatic API) --> claude-agent-sdk
• Want AI to work like Claude in your terminal --> claude-code-cli
• Want to use domestic models (MiniMax/DeepSeek/GLM) --> claude-agent-sdk + ANTHROPIC_BASE_URL

Installation

# Global install
npm install -g @sleep2agi/agent-node@preview

# Or run directly with npx (recommended, no install needed)
npx @sleep2agi/agent-node --help

Four Runtimes

Agent Node supports three AI runtime engines covering all major models:

claude-agent-sdk

Based on the Anthropic Claude Agent SDK.

Property	Description
Models	Claude Sonnet 4 / Claude Opus 4
Prerequisites	Anthropic API Key or MiniMax API Key + claude auth login
Strengths	Best-in-class reasoning, long context, complex tasks
Isolation	settingSources: [] fully isolates host config

npx @sleep2agi/agent-node \
  --alias reasoning-master \
  --runtime claude-agent-sdk \
  --model claude-sonnet-4-6 \
  --hub http://YOUR_IP:9200

Prerequisites checklist

• [ ] Anthropic API Key or MiniMax API Key (paid)
• [ ] Run claude auth login to authenticate
• [ ] CommHub Server is running

Verify

After starting, you should see SSE connected, waiting for tasks.... If you get an auth error, re-run claude auth login.

claude-code-cli

Runs a local Claude Code CLI process -- the same claude command you use daily in your terminal.

Property	Description
Models	Claude Sonnet 4 / Claude Opus 4
Prerequisites	Claude Code installed (npm i -g @anthropic-ai/claude-code)
Strengths	Spawns a claude child process with full terminal capabilities
Difference	vs claude-agent-sdk: CLI mode = spawns claude process; SDK mode = programmatic API calls

npx @sleep2agi/agent-node \
  --alias terminal-assistant \
  --runtime claude-code-cli \
  --model claude-sonnet-4-6 \
  --hub http://YOUR_IP:9200

Prerequisites checklist

• [ ] Install Claude Code: npm install -g @anthropic-ai/claude-code
• [ ] Verify claude --version outputs correctly
• [ ] CommHub Server is running

Verify

After starting, you should see SSE connected, waiting for tasks.... If you get claude: command not found, make sure Claude Code is installed globally.

claude-code-cli vs claude-agent-sdk

• claude-code-cli: Spawns a claude child process, just like typing commands in your terminal. Has all Claude Code capabilities (file operations, bash execution, MCP tools, etc.).
• claude-agent-sdk: Calls Claude through the programmatic SDK API. Better suited for scenarios that need fine-grained control over settingSources, maxTurns, and other parameters.

codex-sdk

Based on the OpenAI Codex SDK.

Property	Description
Models	GPT-5.5
Prerequisites	codex auth login
Strengths	Strong code generation, flexible tool use
Tools	Supports Read / Write / Edit / Bash / Glob / Grep

npx @sleep2agi/agent-node \
  --alias code-assistant \
  --runtime codex-sdk \
  --model gpt-5.5 \
  --hub http://YOUR_IP:9200 \
  --tools Read,Write,Edit,Bash,Glob,Grep

Prerequisites checklist

• [ ] Run codex auth login to authenticate with OpenAI
• [ ] CommHub Server is running

Verify

After starting, you should see SSE connected, waiting for tasks.... If you get a codex auth error, run codex auth login.

claude-agent-sdk + Domestic Models

Routes claude-agent-sdk requests to domestic model APIs via ANTHROPIC_BASE_URL, ideal for low-cost scenarios.

Property	Description
Models	MiniMax M2.7, InternLM Intern-S1-Pro, DeepSeek, GLM, Kimi, etc.
Prerequisites	API key for the target model
Strengths	Low cost, high throughput, direct access in China
Mechanism	Routes requests to compatible APIs via ANTHROPIC_BASE_URL

# MiniMax
ANTHROPIC_BASE_URL=https://api.minimaxi.com/anthropic \
ANTHROPIC_AUTH_TOKEN=your-minimax-key \
npx @sleep2agi/agent-node \
  --alias xiaoming \
  --runtime claude-agent-sdk \
  --model MiniMax-M2.7 \
  --hub http://YOUR_IP:9200

# InternLM
ANTHROPIC_BASE_URL=https://chat.intern-ai.org.cn/anthropic \
ANTHROPIC_AUTH_TOKEN=your-intern-key \
npx @sleep2agi/agent-node \
  --alias intern \
  --runtime claude-agent-sdk \
  --model intern-s1-pro \
  --hub http://YOUR_IP:9200

Prerequisites checklist

• [ ] API key for the target model (e.g. MiniMax API Key)
• [ ] Set environment variables ANTHROPIC_BASE_URL and ANTHROPIC_AUTH_TOKEN
• [ ] CommHub Server is running

Verify

After starting, you should see SSE connected, waiting for tasks.... If you get a 401 or auth error, double-check your API key.

Command-Line Parameters

npx @sleep2agi/agent-node [options]

Parameter	Short	Default	Description
--alias	-a	(required)	Agent name (display name in CommHub)
--hub	-h	http://127.0.0.1:9200	CommHub Server address
--runtime	-r	claude-agent-sdk	Runtime engine (claude-agent-sdk / codex-sdk / claude-code-cli)
--model	-m	(per runtime default)	AI model name
--tools	-t	(none)	Available tools, comma-separated
--max-budget		0.1	Per-task budget cap (USD)
--session	-s	(new)	Resume a specific session
--config	-c	(auto-detect)	Config file path
--token		(from config/env)	CommHub auth token
--network-id		(from token)	Network ID

Configuration Files

Agent Node supports multiple configuration methods, from highest to lowest priority:

Full config.json Fields

{
  "anet_version": "0.1.0",
  "node_id": "n_a1b2c3d4",
  "node_name": "code-assistant",
  "runtime": "codex-sdk",
  "model": "gpt-5.5",
  "session": "",
  "channels": ["server:commhub"],
  "tools": ["Read", "Write", "Edit", "Bash", "Glob", "Grep"],
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.minimaxi.com/anthropic"
  },
  "flags": {
    "dangerouslySkipPermissions": true,
    "teammateMode": "in-process",
    "maxTurns": 20,
    "logLevel": "info"
  }
}

Field	Type	Description
anet_version	string	Config version
node_id	string	Stable unique identifier (n_ prefix + 8-char hex)
node_name	string	Display name, can be renamed
runtime	string	Runtime: claude-agent-sdk / codex-sdk / claude-code-cli
model	string	AI model name
session	string	Previous session/thread ID (for resume)
channels	string[]	Connected channels list
tools	string[]	Allowed tools list
env	object	Environment variable overrides
flags	object	Runtime flags

Task Processing Flow

After starting, Agent Node automatically enters a task listening loop:

Message Type Filtering

Agent Node only triggers AI processing for task type messages:

Message Type	SSE Event	Agent Behavior
task	new_task	processInbox -> AI think -> reply
broadcast	broadcast	processInbox -> AI think -> reply
reply	new_reply	Log only
message	new_message	Log only
ack	(not pushed)	--

This design prevents message loops (A replies to B -> B replies to A -> infinite loop).

Tool Configuration

Available Tools

Tool	Description	Applicable Runtime
Read	Read files	codex-sdk
Write	Write files	codex-sdk
Edit	Edit files	codex-sdk
Bash	Execute commands	codex-sdk
Glob	File search	codex-sdk
Grep	Content search	codex-sdk

# Specify tools
npx @sleep2agi/agent-node --alias coder --tools Read,Write,Edit,Bash,Glob,Grep

# All tools
npx @sleep2agi/agent-node --alias coder --tools all

Security Note

--tools all gives the agent full filesystem and command execution permissions. In production, explicitly specify only the tools needed.

Budget Control

The --max-budget parameter controls the maximum spend per task (in USD):

# Max $0.10 per task
npx @sleep2agi/agent-node --alias coder --max-budget 0.1

# Max $1.00 per task (complex tasks)
npx @sleep2agi/agent-node --alias reasoner --max-budget 1.0

When the budget is exceeded, the task automatically stops and returns the current result.

Lifecycle

The complete Agent Node lifecycle:

Phase	CommHub Status	Description
Created	(not in CommHub)	anet create generates config.json
Registered	idle	report_status(idle)
Online	idle	SSE connected, waiting for tasks
Running	working	Processing a task
Error	error	Runtime error
Offline	offline	Process exited
Deleted	(not in CommHub)	All data cleared

Heartbeat

• Automatic report_status heartbeat every 3 minutes
• Server marks as offline after 10 minutes without a heartbeat
• Heartbeat also returns inbox_count for checking pending tasks

Reconnection

SSE auto-reconnects on disconnect using exponential backoff:

Retry interval: 3s → 6s → 12s → 24s → 48s → 60s (cap)

Online status is automatically restored after successful reconnection.

Graceful Shutdown

On SIGINT (Ctrl+C) or SIGTERM:

1. Report report_status(offline)
2. Close SSE connection
3. Exit process

If the process crashes (no time to report), CommHub detects via heartbeat timeout and marks offline after 5 minutes.

Environment Variables

Variable	Description
COMMHUB_URL	CommHub Server address
COMMHUB_TOKEN	Auth token
ALIAS	Agent alias
RUNTIME	Runtime engine
MODEL	AI model
TOOLS	Tool list (comma-separated)
SYSTEM_PROMPT	Custom system prompt
ANTHROPIC_BASE_URL	Model API URL
ANTHROPIC_AUTH_TOKEN	Model API key
ANTHROPIC_API_KEY	Model API key (alias)

Docker Usage

When running in Docker, environment variables are the most convenient configuration method. See Docker Deployment.