Documentation

Applies to SDK 0.16+ · Last updated: 2026-06-12

AgentNode is the package manager and secure runtime for AI agent skills: agents discover, install, and run verified tools at runtime. This documentation covers the CLI, Python SDK, and REST API — from first install to publishing your own packages.

terminal

$ pip install agentnode-sdk
$ agentnode setup
$ agentnode install word-counter-pack

New here? Start with the Quick Start.

Start

Quick Start

Zero to your first verified agent skill in minutes.

Installation

Requirements, install options, and CLI verification.

Search & Discovery

How search and the resolution engine find the right pack.

Installing Packages

Install flow, hash verification, lockfile, updates.

Runtime & Security

LLM Runtime

Connect OpenAI, Anthropic, and Gemini agents to the registry.

LLM Providers

OpenRouter, DeepSeek, Mistral, Qwen, Gemini, local Ollama.

Agents

Agent packages: manifests, tiers, tool access.

Execution Sandbox

Community code runs isolated — or not at all.

Security Model

What is enforced, per trust tier — honestly documented.

Credentials & Connectors

Connector tokens, OAuth, and credential resolution.

AgentNode Guard

Pre-execution policy gateway: allow, prompt, deny, audit.

Trust & Security

Four trust levels, signatures, and the permission model.

Data Sovereignty

What stays local and what the registry sees.

Publishing & Registry

Publishing Guide

From project structure to a published, verified pack.

Package Verification

The verification pipeline, scoring, and tiers.

ANP Manifest Reference

Full ANP manifest field reference.

GitHub Action

Publish packs from CI with the official action.

Import Tools

Convert MCP, LangChain, and OpenAI tools into packs.

Reference

CLI Reference

Every CLI command with flags and examples.

Python SDK

The Python SDK surface: search, resolve, install, run.

REST API

REST endpoints with request/response examples.

MCP Integration

Use AgentNode through the Model Context Protocol.

Also see the changelog, GitHub repository, and llms.txt for AI tooling.