Skip to main content
Every AI agent eventually needs to do something — run code, edit files, call an API, keep state between runs. Runtools is the infrastructure for that part: private cloud sandboxes, registered local devices, persistent workspaces, hosted tools with server-side credentials, and durable thread history. Give your agent a place to work, not just a place to talk. Drive it visually from the dashboard, or wire it into your own product with the SDK, CLI, and REST API — same platform underneath.

Sandboxes

Create isolated Linux or desktop environments, run commands, expose dev servers, pause idle work, and connect over SSH.

Workspaces

Persist project files, mount them into sandboxes, and give sandbox-backed agents filesystem memory across runs.

Agents

Define model, prompt, tools, execution mode, and resources. Run agents from the dashboard, CLI, SDK, REST API, or a registered device.

Tool Hub

Install public tools, deploy org-private tools with defineTool(), and let Runtools resolve OAuth or stored credentials server-side.

Product Apps

Build customer-facing agents where each end user connects their own integrations through Hosted Connect.

RunMesh

Register devices and private nodes so local agents can run on the right machine while still writing to cloud threads.

Current Public Surface

Quick Example

Where To Start

Quickstart

Create an API key, run the CLI, and create a sandbox.

Core Concepts

Understand sandboxes, agents, tools, projects, and credentials.

Product Apps

Build SaaS-style products where your own users connect their own tools.

Hosted Connect

Add end-user OAuth and API-key integrations to your product.

API Reference

Use the REST APIs directly.

SDK

Use the TypeScript SDK in your application.

Threads

Cloud thread history, live WebSocket frames, compaction, and local runtime event publishing.