Action Gateway Quickstartpublic
Last verified 22 Oct 2026
Action Gateway connects applications and AI agents to tools through a managed MCP endpoint or SDK, with the accounts and permissions you choose. See What You Can Build for standalone agent examples.
Give your coding agent access to Action Gateway tools by adding one MCP server URL. Create the URL in the Control Panel, connect your agent, and try a read-only tool call.
This quickstart connects Claude Code or Codex in three steps. You need a DigitalOcean account and either client installed. For clients such as OpenClaw, Hermes, or ChatGPT, see Connect an External MCP Client.
1. Get Your MCP URL
Action Gateway calls the saved configuration behind this URL a session. It controls tool access, not your agent’s chat history or workspace.
- Open Action Gateway Sessions in the Control Panel and click Create session.
- Enter a Session Name, such as
coding-agent, and keep the default actor ID and other default settings. The default Ask permission policy requires your approval before tools run. - Create the session and copy its MCP URL. This is the server URL you add to your coding agent in the next step.
2. Connect Your Coding Agent
Choose your agent below. Run the commands in a terminal from your project directory, replacing <action-gateway-session-mcp-url> with the URL you copied. When prompted, sign in with the DigitalOcean account you used to create the session.
Add Action Gateway and start browser sign-in:
codex mcp add action-gateway --url "<action-gateway-session-mcp-url>"
codex mcp login action-gatewayAfter signing in, run codex to start a new coding-agent session.
Add Action Gateway:
claude mcp add --transport http action-gateway "<action-gateway-session-mcp-url>"Run claude in the same project directory. In Claude Code, run /mcp, select action-gateway, and follow the browser sign-in prompts.
3. Try a Tool
Send this prompt to your coding agent:
Use Action Gateway to list available DigitalOcean Droplet sizes. Do not create, update, or delete any resources.Review and approve the read-only call when prompted. Your agent finds the tool, runs it, and returns a list of Droplet sizes.
This example does not require a prepaid balance, but tool invocation charges apply.
Reuse this MCP URL for future tasks. To change its tools or permissions, create another session and update your client’s URL.
Next Steps
- To work with your GitHub repositories or Linear issues, connect those accounts using the session’s actor ID. Keep the same MCP URL.
- Before deployment, configure Action Gateway for production agents to restrict tools and permissions.
- Try a worked example, such as Jira issue triage, or explore What You Can Build.
- To connect your own application, use the SDK setup examples.