Node Runtime

Every Agent Node has a Runtime (engine kernel) that decides how the node calls models and runs tools. Agent Network ships three Runtimes — and you can mix them on a single Hub: a Claude Code CLI agent dispatches a translation task to a MiniMax agent, then asks a Codex agent to write code, and merges the results back.

Three runtimes at a glance

Runtime	Engine	When to pick	Default models	Auth
`claude-code-cli`	spawn local `claude` CLI	Reuse your Claude subscription	Claude Sonnet / Opus	`claude` already logged in
`claude-agent-sdk`	`@anthropic-ai/claude-agent-sdk`	Programmatic access to any Anthropic-compatible API	Anthropic / MiniMax / DeepSeek / GLM / Kimi	API Key
`codex-sdk`	`@openai/codex-sdk`	Code writing / shell commands	OpenAI Codex (gpt-5)	`codex auth login`

Not sure which one?

Want to reuse a Claude subscription → claude-code-cli
Writing copy / translation / analysis (programmatic) → claude-agent-sdk
Writing code / running commands → codex-sdk
Using MiniMax / DeepSeek / GLM / Kimi → claude-agent-sdk + ANTHROPIC_BASE_URL
Mix-and-match (recommended) → run all three on one Hub, pick the best engine per role

claude-code-cli

Reuses your already-logged-in Claude CLI session — no API key, no token, just works.

How it works

anet node start  →  spawn `claude` subprocess
                 ↓
         commhub MCP (stdio) tools injected
                 ↓
         receive send_task → forward to Claude session → reply

The node spawns a claude CLI subprocess on start
The commhub MCP server (over stdio) injects "receive task / dispatch task / reply" tools into the Claude session
When a task arrives, the Hub pushes via SSE → MCP forwards to Claude → Claude processes and replies

When to pick

You already use Claude Code (claude.ai subscription)
You want to plug your daily Claude session into multi-agent collaboration
You don't want to pay for API access separately

Config

bash

anet node create my-bot --runtime claude-code-cli
anet node start my-bot

config.json:

json

{
  "runtime": "claude-code-cli",
  "session": "",
  "flags": {
    "dangerouslySkipPermissions": true,
    "teammateMode": "in-process"
  }
}

Notes

Requires a working claude --version (Claude Code CLI installed and authenticated)
session is auto-saved; you can anet node resume later to keep the conversation
Key difference vs the SDK runtime: CLI gives you the full Claude Code toolset (file ops / Bash / MCP)

claude-agent-sdk

Programmatic access to any Anthropic-compatible API — Anthropic by default, but redirectable to MiniMax / DeepSeek / GLM / Kimi via ANTHROPIC_BASE_URL.

How it works

anet node start  →  spawn agent-node subprocess
                 ↓
         @anthropic-ai/claude-agent-sdk
                 ↓
         POST → ANTHROPIC_BASE_URL (default api.anthropic.com)
                 ↓
         commhub MCP (stdio) over SSE

The agent-node process drives the SDK to call any Anthropic-compatible API
Override the base URL with ANTHROPIC_BASE_URL to redirect to any compatible provider
settingSources: [] fully isolates the agent from your local ~/.claude/ config

When to pick

Direct Anthropic API (no subscription required)
Domestic Chinese models (MiniMax / DeepSeek / GLM / Kimi — cheap / high-throughput / fast)
Per-task model switching

Config

Anthropic direct:

bash

ANTHROPIC_API_KEY=sk-ant-xxx \
anet node create planner \
  --runtime claude-agent-sdk \
  --model claude-sonnet-4-6

MiniMax:

bash

anet node create translator \
  --runtime claude-agent-sdk \
  --model MiniMax-M2.7 \
  --env "ANTHROPIC_BASE_URL=https://api.minimaxi.com/anthropic" \
  --env "ANTHROPIC_AUTH_TOKEN=sk-cp-xxx"

Verified models

Provider	Model	`ANTHROPIC_BASE_URL`
Anthropic	claude-sonnet-4-6 / claude-opus-4-7 / claude-haiku-4-5	`https://api.anthropic.com`
MiniMax	MiniMax-M2.7	`https://api.minimaxi.com/anthropic`
DeepSeek	deepseek-v3-0324	(per provider docs)
Zhipu GLM	glm-4-plus	(per provider docs)
Moonshot Kimi	kimi-k1.5	(per provider docs)
InternLM	intern-s1-pro	`https://chat.intern-ai.org.cn/anthropic`

codex-sdk

OpenAI Codex CLI runtime — best for writing code and running shell commands.

How it works

anet node start  →  spawn @openai/codex-sdk
                 ↓
         OpenAI API (OPENAI_API_KEY)
                 ↓
         commhub MCP (stdio) over SSE

Driven by the official @openai/codex-sdk package
Supports Read / Write / Edit / Bash / Glob / Grep tools
Auth via codex auth login (OAuth) or OPENAI_API_KEY

When to pick

OpenAI's official Codex / latest gpt-5
Code generation / shell command execution
Heavy tool calling / function calling

Config

bash

codex auth login  # one-time

anet node create coder \
  --runtime codex-sdk \
  --model gpt-5 \
  --tools Read,Write,Edit,Bash,Glob,Grep

Verification status

codex-sdk passes unit tests but lacks full end-to-end coverage (real codex auth regressions are missing). For production runs, smoke-test with a trivial task first.

Cross-runtime mesh

The core value: dispatch tasks across runtimes on the same Hub.

bash

# 1. Claude Code CLI agent — planner using your subscription
anet node create planner --runtime claude-code-cli

# 2. MiniMax agent — translation
anet node create translator \
  --runtime claude-agent-sdk \
  --model MiniMax-M2.7 \
  --env "ANTHROPIC_BASE_URL=https://api.minimaxi.com/anthropic" \
  --env "ANTHROPIC_AUTH_TOKEN=sk-cp-xxx"

# 3. Codex agent — code writing
anet node create coder --runtime codex-sdk --model gpt-5

# 4. Start all three
anet node start planner
anet node start translator
anet node start coder

In the Dashboard, ask planner:

Translate this English passage to Chinese, then have coder write a Python script that writes the result to a file.

planner will use commhub MCP tools:

get_all_status — sees translator + coder online
send_task(alias="translator", task="translate ...")
get_task — polls for the translation
send_task(alias="coder", task="write a script ...")
Combines both results and replies to you

The whole flow is visible in real time on the Tasks / Messages dashboard pages.

Cheat sheet

Goal	Pick
Already paying for Claude, no API budget	`claude-code-cli`
Domestic Chinese model (MiniMax / DeepSeek / GLM / Kimi)	`claude-agent-sdk` + `ANTHROPIC_BASE_URL`
Anthropic official API (stable backend)	`claude-agent-sdk`
Code writing / shell	`codex-sdk`
Copy / translation / analysis / RAG	`claude-agent-sdk`
Full Claude Code toolset (file / Bash / MCP)	`claude-code-cli`
Team mesh (planner + translator + coder)	All three, pick per role

Verified vs not

Verified (v2.1 E2E)

claude-agent-sdk — Anthropic / MiniMax / DeepSeek / GLM / Kimi / InternLM all pass E2E
Multi-runtime mesh (peer agents auto-coordinate via get_all_status + send_task + get_task)

Not verified

claude-code-cli — runs locally, no E2E regression
codex-sdk — unit-tested only, real codex auth E2E pending

Node Runtime ​

Three runtimes at a glance ​

claude-code-cli ​

How it works ​

When to pick ​

Config ​

Notes ​

claude-agent-sdk ​

How it works ​

When to pick ​

Config ​

Verified models ​

codex-sdk ​

How it works ​

When to pick ​

Config ​

Cross-runtime mesh ​

Cheat sheet ​

Verified vs not ​

Next ​

Node Runtime

Three runtimes at a glance

claude-code-cli

How it works

When to pick

Config

Notes

claude-agent-sdk

How it works

When to pick

Config

Verified models

codex-sdk

How it works

When to pick

Config

Cross-runtime mesh

Cheat sheet

Verified vs not

Next