Skip to content
For developers building with Hermes

Managed AI agent hosting for developers

Move a useful agent workflow onto a hosted Hermes runtime. cloudrun provisions the box and gives you a canvas for chat, configuration, tools, integrations and scheduled work.

Start with the workflow you can already explain and test. Connect its dependencies, inspect the tool calls and prove the result on the hosted setup.

Hermes available today. View hosting pricing

What you can configure

Keep the runtime and its configuration within reach

Start from a prepared runtime image

Creating a project restores a prebuilt Hermes installation and opens the project's canvas. The installation image prepares the runtime; recovery of your own state needs a separate plan.

Inspect a task as it runs

The agent's Overview tab provides streaming chat with tool calls shown inline. Use a representative request to see which tools it invokes and what it returns.

Connect your workflow's interfaces

Use Tools for toolsets and MCP servers, Skills for instruction packs and Integrations for supported connected services.

Choose your model provider

The documented Models panel lets you connect your own supported provider. Record credential ownership and usage expectations as part of the deployment.

From setup to useful work

From project creation to a tested task

Follow the product quickstart
  1. Create an organization and project

    The organization is your workspace. Creating a project provisions its first Hermes engine and default agent.

  2. Open the agent on the canvas

    Wait for the runtime to be ready, then open the agent's Overview tab.

  3. Configure the dependencies

    Add the instructions, required tools and integrations. Verify the files and services your task needs.

  4. Run a representative task

    Use a known expected result and inspect the output and tool activity.

  5. Add delivery and scheduled work

    Add a channel or automation after the interactive task works. Verify the complete delivery path.

A resource you can use today

A clear handoff starts before launch

Record account ownership, permissions, expected results and recovery steps. Use the client-agent launch checklist to decide what is ready and what still needs a test.

Get the launch checklist

Reusable worksheet. No signup required.

Make the next decision

Make the hosting decision around the workload

List the files, executables, credentials and external services your workflow needs. Check those dependencies on the destination before switching over an existing agent. Plan the state transfer, channel cutover and rollback before moving live traffic.

Before you get started

Questions developers ask

Which runtimes can I deploy?

Hermes is the currently supported runtime. OpenClaw, Codex, Claude Code, OpenCode, Pi and Cursor are marked coming soon. Confirm support for your intended runtime before planning a deployment.

Read the engine documentation
Can I deploy any arbitrary agent application here?

This page describes the supported Hermes workflow. Confirm compatibility before treating cloudrun as a host for a different runtime, custom service or framework.

Can I bring a model-provider API key?

The Models panel supports connecting your own providers. The inspected credential path passes the key through the control plane to the runtime; it is not a zero-knowledge key flow.

Understand BYOK access and revocation
Can I migrate an existing local agent?

Inventory its state, dependencies and credentials first, then verify which import and export steps the destination supports. Moving between runtimes requires its own evaluation.

Use the migration checklist
Are Hermes profiles separate security boundaries?

Profiles organize agents within the runtime. Do not use profile separation alone as proof of isolation for mutually untrusted workloads.

Review the isolation boundaries

Your first project

Test your workflow on a hosted runtime

Create a Hermes project and run a task with a known result. Connect the rest of the workflow once that first check passes.