Cloud Agents

AdaL Cloud Agents is a hosted agent runtime. You define an agent (model + system prompt + tools), AdaL runs the harness and infrastructure, and you interact with it over HTTP — as if it were any other agents API.

Instead of building your own agent loop, tool execution, and runtime, you get a fully managed environment where the agent can call your custom tools, maintain state across turns, and stream responses back to your app.

Two ways to build with AdaL

	AdaL CLI / SDK	AdaL Cloud Agents
What it is	The agent runtime on your machine or in your Python app	A hosted agent harness running in managed cloud infrastructure
Best for	Local development, coding tasks, fine-grained control	Production agents, long-running workloads, non-Python apps
Infrastructure	You run it	AdaL runs it for you
Access	Terminal or Python SDK	HTTP API from any language
Learn more	CLI Quickstart · SDK	This section

Core concepts

Cloud Agents is built around three concepts:

Concept	Description
Agent	A reusable configuration: a name, a system prompt, optional custom tools, and which built-in tools it can use. Create it once, reference it by ID across sessions.
Session	A running agent instance performing a specific task. Sessions are stateful — conversation history persists, and you can resume them later.
Events	Messages exchanged between your app and the agent: user turns, tool calls and results, and streamed assistant responses.

How it works

1. Create an agent — define the model, system prompt, and tools. The agent is stored in the cloud and referenced by ID.
2. Start a session — launch a session that references your agent. Pick a model; AdaL provisions a persistent agent worker.
3. Send events and stream responses — send user messages over HTTP. The agent autonomously calls tools and streams back results via server-sent events.
4. Steer or resume — send additional messages to guide the agent mid-task, or resume a paused session later. Event history is persisted server-side.

Your app  ──HTTP──▶  AdaL Cloud  ──▶  Agent (prompt + tools + model)
                         │
                         └── streams responses + tool events back

When to use Cloud Agents

Cloud Agents is best for workloads that need:

• No infrastructure — you don't want to build or maintain an agent loop, sandbox, or tool-execution layer.
• Production agents — agents serving real users, where uptime and scaling matter.
• Non-Python apps — any language with HTTP can call the agents API (web, mobile backend, CI script).
• Shared agents — one agent definition used across multiple frontends or services.
• Stateful sessions — persistent conversation history and resumable sessions across interactions.

If you'd rather embed AdaL directly in a Python app, use the SDK. If you want the terminal coding agent, use the CLI.

Supported tools

Cloud agents can use:

• Built-in tools — file operations, bash, search, web access, image/video generation, and multi-model consult. Control which ones the agent sees via --enabled-default-tools / --disabled-default-tools.
• Custom tools — Python functions in .adal/tools.py that the agent can call (catalog lookups, quote builders, lead capture, anything you can write).
• MCP servers — connect to external tool providers.

Next

• Quickstart — create your first cloud agent
• Live Demo — a customer sales agent running on AdaL Cloud