Send it a prompt

The Agent API on each template takes a message and streams the reply; sessions continue a thread.

docs-confirmed “Send a message or task to the agent on your instance.” Source↗

One call queues a prompt to a named harness, with steering, interrupts and an event stream.

docs-confirmed “Queue a natural-language work item for an agent harness” Source↗

A prompt call returns a prompt id and an event stream, with steering for three harnesses.

docs-confirmed “The call returns at once with `promptId`, `conversationId` and `status: "queued"`.” Source↗

The catch: You can send a prompt from the command line, but the SDK and API have no prompt call, so a program cannot start an agent.

docs-confirmed “`lizard sandbox prompt <id> <prompt...>` sends a prompt and streams what the agent does.” Source↗

Note: There is no prompt endpoint; Agent Runtime, the product that would run the loop, is marked Soon.

Agents Hosting runs your own deployed agent, not a prompt-to-work API.

docs-confirmed “Session-first, stateful runtime for long-running agents.” Source↗

Note: boxd runs the machine, not the agent loop; Anthropic or OpenAI runs the loop in the documented integrations.

docs-confirmed “Anthropic runs the agent loop.” Source↗

Note: Your Worker holds the credentials and the agent runs as a process you start; there is no prompt endpoint.

docs-confirmed “Your Worker starts the sandbox, holds the credentials, and decides which hostnames the sandbox can reach.” Source↗

Note: No agent is hosted and there is no prompt call; you install and drive the agent.

docs-confirmed “Run Claude Code in a Daytona Sandbox via CLI” Source↗

Note: Guides show agents running in a sandbox, but no API takes a prompt and runs one for you.

docs-confirmed “Run AI coding agents like Claude Code, Codex, and Amp in secure E2B sandboxes with full terminal, filesystem, and git access.” Source↗

Note: There is no prompt endpoint; the API offers exec, files, desktop and snapshots.

Your agent runs as a process you start inside the machine.

inferred “exec`: Run a command” Source↗

Note: Modal sells the computer, not an agent loop; the docs show SDK code you write.

inferred “Sandboxes are useful for tasks like executing code generated by an LLM” Source↗

Note: The tmux API lets your program type into an agent, but Morph runs no agent loop of its own.

docs-confirmed “Devboxes expose a full tmux API, which lets you programmatically manage coding agents like OpenAI Codex or Claude Code.” Source↗

Note: Runtime has no agent-run API; its own docs say no bridge harness has run on it yet.

docs-confirmed “No bridge harness (Claude Code, Codex) has run on Runtime yet.” Source↗

Note: Sprites host an agent you install; there is no prompt endpoint.

inferred “Give it a Sprite that sleeps until it's needed and wakes up remembering everything.” Source↗

Note: Tensorlake exposes a sandbox as a tool for your own agent; it does not run an agent from a prompt.

inferred “Expose Tensorlake sandboxes as tools to your LLM agents” Source↗

Note: Agents are in the image but there is no API that runs one; you open a terminal or use the CLI.

docs-confirmed “The default image is `vercel/sandbox/universal`, which includes the current Node.js LTS, Python 3.14, coding agents, and common utilities.” Source↗