Introduction
Nod connects your AI agents to the people who work with them
What is Nod?
Nod is a platform that sits between your AI agents and the humans who need to interact with them. It provides a real-time interface for chatting with agents, approving their actions, scheduling tasks, and monitoring everything they do.
Think of it as the communication layer for agentic systems. Your agent connects to Nod via API or WebSocket. Your users interact through the Nod app (mobile, web, desktop). Messages, decisions, and activity flow between them in real time.
How it works
Every integration with Nod starts with an agent — an entity in Nod that represents your external agentic system. When you create an agent in Nod, you get an agent_id and a secret. These credentials let your system authenticate and communicate with Nod.
From there, your agent can:
- Send messages that appear in the user's chat
- Create requests that ask humans for approvals, choices, or answers
- Log activity so users can see what the agent is doing
- Execute tasks on a schedule or triggered by external events
- Receive user messages and decision responses in real time
Users interact with all of this through the Nod app — a clean, mobile-first interface designed specifically for human-agent collaboration.
Agent types
Nod supports two agent types, depending on how you want the interaction to work:
| Type | Description | Best for |
|---|---|---|
| Interactive | Full bidirectional chat. The agent sends messages, receives replies, creates requests, and the user can message the agent freely at any time. | Coding assistants, research agents, general-purpose assistants |
| Approval-only | The agent only creates requests (approvals, choices, questions). No chat, no tasks. Users respond to requests but cannot message the agent. | Workflow automation, approval gates, human-in-the-loop |
Both types have access to the full API — tasks, activity logging, webhooks, etc. The difference is purely in the chat experience shown to users.
Core features
How to integrate
Nod is designed to work with any agentic system. There are three main ways to connect:
| Method | When to use | Details |
|---|---|---|
| WebSocket | Always-running agents that need real-time communication | Persistent connection. Send and receive messages, decisions, activity instantly. |
| REST API | Serverless functions, cron jobs, or simpler integrations | Stateless HTTP calls. Send messages and requests, poll for tasks. |
| Webhooks | Receive events when offline, or trigger tasks from external services | Nod delivers events to your URL. External services trigger tasks via public URLs. |
Most integrations use a combination. For example, our open-source CLI uses WebSocket for real-time communication and REST for task polling. A simpler integration might use only REST and webhooks.
Authentication
Agents authenticate with two headers: their ID and their secret.
| Header | Description |
|---|---|
| X-Nod-Agent-Id | Your agent's unique identifier |
| X-Nod-Secret | Your agent's secret key |
The secret is shown once when you create the agent. The same credentials work for REST, WebSocket, and all authenticated endpoints. See Authentication for examples in multiple languages.