Skip to main content
Use /setup when you want an AI coding agent to add Migma to an existing app. It is agent-guided. The agent still reviews and edits your codebase, then uses Migma to create or reuse emails and wire safe server-side sends. In Migma, the main entry is Create → Set up Migma with agent. That dialog copies the /setup handoff prompt with skill/plugin install paths, sender/domain status, agent auth guidance, and backend wiring steps.
For agents: Fetch the complete documentation index at https://docs.migma.ai/llms.txt to discover all available pages before exploring further.

Use it from Migma

1

Open Create

Click Set up Migma with agent next to the composer send button.
2

Copy for AI

Migma prepares a prompt for Claude Code, Codex, Cursor, or any skills-compatible agent.
3

Paste into your agent

The prompt asks the agent to install the setup skill or plugin only after you approve it.
4

Approve scoped access

The agent uses agent registration first. Use a fallback API key only when agent registration is unavailable.
Use Agent Handoff for the full Create, Publish, API setup, and canvas handoff map.

Add Migma to your agent

Use the plugin when your agent supports it. It installs the setup, migma, and migma-public-api skills together.

Claude Code

Install the plugin:
Or install only the setup skill:
To install across all supported agents on this machine:

Codex

Install the plugin from the terminal:
You can also open /plugins in Codex, search for Migma, and install it from the plugin list after adding the marketplace.

Cursor

Open Cursor → Settings → Plugins, paste https://github.com/MigmaAI/migma-skills, then install Migma. Or install only the setup skill:

Other agents

If your agent supports skills, install the setup workflow:
If it supports MCP but not skills, connect MCP. MCP gives the agent Migma tools, but the setup skill is still the better path for codebase audit and backend wiring instructions.

Terminal install

Use this when you want the CLI and all skills together:
Or install each part:

What setup does

  1. Audits existing email providers, templates, receipts, invites, campaigns, lifecycle jobs, queues, and webhooks.
  2. Resolves the correct Migma brand/project.
  3. Lists existing Migma emails and reuses matches.
  4. Creates missing emails with prompts grounded in trigger files, recipients, variables, subject copy, and CTAs.
  5. Stores each returned emailId, subject, sender, and matching app trigger in server-side config or constants.
  6. Wires sends only in trusted backend paths.
  7. Adds a dry-run or smoke test that verifies each trigger selects the right emailId without sending duplicates.

API access

Use a Migma API key server-side:
Get a key from Settings → Developers → API Keys, or let the agent use agent registration at https://api.migma.ai/auth.md.

REST fallback

If plugin or skill install is unavailable, use REST. Base URL: https://api.migma.ai Auth header: Authorization: Bearer $MIGMA_API_KEY

Guardrails

  • Keep API keys and send calls server-side.
  • Do not put Migma secrets in browser code.
  • Ask before installing packages, calling Migma APIs, creating emails, or sending tests.
  • Test before live sends.
  • Add idempotency before transactional sends.
  • Do not send cold email or email unsubscribed contacts.
  • If the app is frontend-only, stop after audit/email creation and add a backend prerequisite before wiring sends.

Which tool to use

Related docs: Agent Handoff, Skills, CLI, MCP, Authentication, Agent registration.