One remote MCP server. OAuth you approve. Mailboxes, Journeys, Broadcast, Product email, Approvals, and Logs, run by Claude, ChatGPT, Codex, Grok, or the agent you build.
Give your AI agent email on your own domain →
Server URL https://api.bangermail.com/mcp
Choose a guide for your app, with visual setup instructions and a first prompt to try. Developer tools and CLI clients have their own guide.
Ask ChatGPT to welcome every new customer, follow up at the right moment, and win back the ones drifting away. Banger sends the email from your domain; you stay in the chat.
See the stepsAsk Claude to build your welcome Journey, write your next Broadcast, and answer replies. Banger sends it all from your domain, and you review it right in the conversation.
See the stepsAsk Grok to welcome new customers, win back the quiet ones, and send your next Broadcast. Banger runs the email behind it, and every message gets more personal.
See the stepsHermes keeps working after you close the chat, so your welcome, follow-up, and win-back emails run every day through Banger.
See the stepsLet OpenClaw welcome new customers, follow up, and win back the quiet ones through Banger, right from your agent’s conversations.
See the stepsAsk Replit Agent to welcome new signups, answer support mail, and follow up with users who go quiet. Banger runs it from your own domain while Agent builds the app around it.
See the stepsGive an n8n AI Agent Banger as a tool. It welcomes new signups, answers support mail, and follows up with users who go quiet, from your own domain, inside the workflows you already run.
See the stepsConnect Claude Code, Codex CLI, Cursor, or another MCP client. These instructions are for developer tools; ChatGPT and Claude have their own app guides.
See the stepsMore than a hundred tools, grouped the way the product is. Humans, Banger, and agents all see and change the same workspace state.
Start or resume the same guided setup the app shows, as an interactive card in Claude and ChatGPT: send your first email, learn Broadcasts and Journeys, then connect your domain.
Create hosted mailboxes on your domain or a starter address, read and search threads, triage, reply, and see what is waiting on your team.
Receipts, confirmations, alerts, and account messages triggered by your product, with templates, previews, and delivery status per send.
Build an audience, preview who will receive it, check readiness, schedule or send, and follow the timeline and per-recipient outcome.
Every automated flow, from a one-step welcome to a multi-branch lifecycle: create, preview, enroll contacts, activate through review, and inspect executions.
Contacts, lists, segments, and consent-checked imports, so every send targets people who agreed to hear from you.
Generate and revise on-brand email designs from a starter library, keep the brand footer current, and run A/B experiments with recommended arms.
Add sending domains, get the exact DNS bundle, verify, prove activation with a real round trip, and watch deliverability by lane.
Follow what waits for human review, read the audit trail, and read connected services and incoming webhooks without ever seeing a secret.
The authorization page shows the exact scope set, capped by your role. Workspaces are company boundaries; products are strict email boundaries inside them. Sensitive actions follow your approval policy, and every action lands in Logs.
mail:read Read mailboxes, threads, and messages.
mail:write Create mailboxes, labels, drafts, and triage threads.
mail:send Send mailbox email and Product email.
contacts:read Read contacts, lists, and segments.
contacts:write Create and update contacts, lists, segments, and imports.
campaigns:read Read Broadcasts, Journeys, templates, and designs.
campaigns:write Create and edit Broadcasts, Journeys, templates, and designs.
campaigns:send Schedule or send a Broadcast, activate a Journey.
automation:read Read Journey executions, experiments, and Logs.
automation:execute Run governed actions and experiments.
connections:read See which services and webhooks are connected.
connections:use Query connected services through Banger.
connections:admin Connect, rotate, or remove services.
workspace:admin Onboarding, products, sending domains, API keys, and approvals.
Banger implements the MCP authorization spec end to end: RFC 9728 resource metadata, RFC 8414 server metadata, RFC 7591 dynamic registration, PKCE S256, RFC 8707 resource indicators, and rotating refresh tokens. Redirect URIs may be HTTPS or HTTP loopback, so both hosted clients like Grok and local CLIs like Codex register cleanly. A missing scope is filled from the setup ticket or the role, so clients that omit it still get the reviewed grant.
Protected resource metadata
GET https://api.bangermail.com/.well-known/oauth-protected-resource/mcp Authorization server metadata
GET https://api.bangermail.com/.well-known/oauth-authorization-server Dynamic client registration
POST https://api.bangermail.com/oauth/register (public clients, token_endpoint_auth_method none) Authorization
GET https://api.bangermail.com/oauth/authorize (response_type code, PKCE S256, optional scope and resource) Token
POST https://api.bangermail.com/oauth/token (authorization_code and refresh_token, rotating refresh tokens) Unauthenticated request
401 with WWW-Authenticate: Bearer resource_metadata=... so clients start OAuth instead of assuming success Listed in the official MCP Registry as com.bangermail/banger. Plugin manifests for Claude, Grok, and Codex live in bangermail/banger-plugin.
Straight answers about access, approvals, domains, and keys.
Any client that speaks the Model Context Protocol over Streamable HTTP with OAuth: Claude and Claude Code, ChatGPT and Codex, Grok and Grok Build, Cursor, and agents you build yourself. Banger is listed in the official MCP Registry as com.bangermail/banger, and the same plugin package ships manifests for Claude, Grok, and Codex.
Exactly the workspace you approve on the authorization page, with the scope set shown there and capped by your role. Workspaces are company boundaries and products are strict email boundaries inside them; an agent cannot combine audiences, mailboxes, or sends across products. You can revoke the grant from Banger at any time.
Only within your workspace approval policy. Sending, launching a Broadcast, and activating a Journey are governed actions: where your policy requires review, the agent proposes and a person approves in Banger. Governed mutations use idempotency keys, and every action lands in Logs.
No. Onboarding creates a permanent Banger-hosted mailbox and proves a real send-and-receive round trip before it ever asks about a domain. Adding your own domain later gives you the exact DNS bundle and a verification the agent can run for you.
Yes. Create a workspace API key in Banger with the scopes you need and send it as a Bearer token to the same endpoint. OAuth is the right path for interactive agents because it ties the grant to a person and a workspace.
The setup screen in Banger mints a short-lived, workspace-specific URL of the form https://api.bangermail.com/mcp/setup/<ticket>. It pre-binds the grant to your workspace, your client, and the reviewed scope set so the client needs no consent page. The plain endpoint works too; it just asks you to choose the workspace.
Start with a hosted mailbox and a real round trip, add your domain when you are ready, and let the agent you already use run the rest, with you reviewing what matters.