AI Agent Setup

View markdown

AI Agent Setup

Bento docs are available as browser pages and clean markdown so coding agents can read implementation guidance directly.


Goal

Give your agent enough Bento context to choose the right integration path, read the relevant SDK README, and avoid common mistakes around credentials, consent, and deliverability.


Install Bento Skills

If your agent supports skills, install the official Bento skill package first:

Shell
npx skills add bentonow/bento-skills

Bento Skills tells the agent when to use Bento MCP, the CLI, an SDK, or n8n. It also carries safety guidance for API keys, bulk subscriber imports, live account operations, and framework-native setup.

Use Bento Skills for implementation guidance. Use Bento MCP when the agent needs to inspect or operate on a live Bento account from chat.

Bento hosts the MCP server at https://mcp.bentonow.com/mcp, so there's nothing to install. The Bento MCP setup guides cover Claude, Cursor, VS Code, Codex, Gemini CLI, and other AI apps.


Agent Prompt

Use this prompt in Cursor, Claude, Codex, or another coding agent:

Text
Use the Bento skill if it is installed. Then read https://bentonow.com/agent_onboarding.md and talk with me about how we should implement Bento. Follow the docs/SDK links based on what platform we use, and prefer the .md version of Bento docs pages when reading documentation directly.
The docs sidebar includes a copyable version of this prompt.

Machine-Readable Docs

Useful entrypoints:


Implementation Guardrails

Tell your agent:

  • Ask whether the goal is marketing automation or transactional email.
  • Use SDKs when a first-party SDK exists.
  • Keep Bento secret keys server-side.
  • Do not send real production email from local development or CI.
  • Add rate limits to actions that send email.
  • Set up sender authentication before production sending.
Agents should read the relevant SDK README before writing code. SDK method names differ across languages.