Skip to main content

Documentation Index

Fetch the complete documentation index at: https://agent37.com/docs/llms.txt

Use this file to discover all available pages before exploring further.

Each agent is an instance: its own always-on computer with a filesystem, a terminal, a desktop, memory, and connected accounts. Send it a message and it browses, runs code, uses tools, and works on a schedule. There are no servers or containers for you to manage. You hold the key; your end users see only your product.

​
Get started in two calls

Create an agent, then send it a message. You are talking to it in two calls, and the returned session_id continues the conversation from there. 
# 1. create an agent. it comes back running.
curl -X POST https://api.agent37.com/v1/instances \
  -H "Authorization: Bearer sk_live_..."
# -> { "id": "inst_x7", "status": "running", ... }

# 2. talk to it, streamed.
curl -N https://api.agent37.com/v1/responses \
  -H "Authorization: Bearer sk_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "instance_id": "inst_x7",
    "input": "Research the top 3 EV makers, write a memo.",
    "stream": true
  }'

Quickstart

The same two calls, step by step, with prerequisites.

Core concepts

Instances, sessions, and responses, and why the agent is not the model.

​
How it fits together

Three resources, and one distinction worth holding onto.
ResourceWhat it is
Instance inst_The agent’s own always-on computer. Built from a template, persistent until you delete it. Create one per end user.
Session sess_One conversation on an instance. A message starts one; reuse its session_id to continue. An instance can hold many.
Response resp_One agentic turn: your input, the agent’s work, its reply. Stream it live or fetch it by id.
The agent is not the model. The template installs the agents (Hermes, Claude Code, Codex, OpenClaw, or a mix); a session runs one of them. The model is the LLM that agent runs on, chosen per message as model plus provider. They are separate dials.

​
Building blocks

Snap these together to build your product. Every one acts on an instance.

Send a message

The core call: input in, an agentic reply out. Continue and manage conversations.

Streaming events

Render text, reasoning, and tool activity live over Server-Sent Events.

Skills & plugins

Preinstall skills and add new tools so the agent reaches for them on its own.

Files

Put files on the agent’s filesystem, then ask about them by path.

Run commands

Run any shell command inside the instance from your backend.

Integrations

Connect Gmail, Slack, Notion, and 1000+ apps. Routed and metered for you.

Scheduled runs

Give an agent a cron schedule; it runs unattended and you read the results.

Terminal, desktop & files

Embed a live terminal, desktop, or file browser with signed URLs.

​
Guides

Build a chat app

Give every user their own always-on agent. Where most teams start.

​
Building with an AI coding agent

Point your coding agent at the machine-readable API map and it can scaffold a working client: 
https://api.agent37.com/llms.txt
Looking for OpenClaw: channels, model providers, and self-hosting? Switch products in the selector at the top of the sidebar, or start at the OpenClaw overview.