Features

🛡️ Local server
- Email contents, custom templates, recipient groups and sender identities are stored only on your machine.
✉️ Send emails in plain text, HTML or Markdown to one or more recipients.
✅ Send consentless transactional emails using built-in templates (OTP, MFA, Order Completion, etc.)
🤝 Automatic consent flow handling.
- Recipients added via MCP receive the consent email.
- Emails are queued locally until recipient grants the consent to receive them.
📝 Create, edit and store custom templates with dynamic variables.
- Interactive template previews for quick design iteration.
- Send test email to verify template rendering.
👥 Organize recipients into groups and reference them all by group name.
🏷️ Manage sender identities (From: header) and reference them by name.
📊 Monitor email health metrics.
📰 Maintain the blog: settings, custom domain, drafts, publish, preview, and email broadcast.

Installation

MCP Bundle (`.mcpb` file, easiest)

Grab the goodsender.mcpb file from the latest release.
Double-click the goodsender.mcpb file to open it in your AI client.

Note: Not every AI client supports MCP bundles or local MCP servers yet. Use a different installation method if nothing happens on double-click.

Docker image / self-hosting

Run via

docker run -d -e GOODSENDER_API_KEY=<Your API Key> -v goodsender-mcp-data:/data ghcr.io/good-sender/mcp:latest

If your goal is to run the agent using this MCP fully autonomously, set GOODSENDER_DISABLE_ACTION_GUARDS=true to skip all action confirmations completely. See Action confirmation.

or add to docker-compose.yaml on your home server

services:
  goodsender-mcp:
    image: ghcr.io/good-sender/mcp:latest
    ports:
      - "9889:9889"
    environment:
      GOODSENDER_API_KEY: <Your API Key>
      # GOODSENDER_DISABLE_ACTION_GUARDS: "true"  # skip all action confirmations
    volumes:
      - goodsender-mcp-data:/data
    restart: unless-stopped

volumes:
  goodsender-mcp-data:

and run via

docker compose up -d

Manual configuration (JSON)

Grab the binaries.zip file from the latest release (includes MCP goodsender-mcp-* and CLI goodsender-* binaries).
Unzip it somewhere you can easily reference.
Modify your AI client config file, adding the MCP server configuration (example for Claude Desktop and Cursor on macOS):
```
{
  ...
  "mcpServers": {
    "GoodSender": {
      "command": "<path to the unzipped binaries>/goodsender-mcp-darwin-arm64",
      "env": {
        "GOODSENDER_API_KEY": "<Your API key>"
      }
    }
  }
  ...
}
```
For other platforms you must reference a corresponding binary instead of darwin (macOS)

If your goal is to run the agent using this MCP fully autonomously, set GOODSENDER_DISABLE_ACTION_GUARDS=true to skip all action confirmations completely. See Action confirmation.

Action confirmation

By default, mutating tools (create, publish, send, and similar) do not run immediately. They return a confirmation token. The agent must call confirm_action after you approve.

Set GOODSENDER_DISABLE_ACTION_GUARDS=true in the MCP server environment to skip that step. Tools then run when the client calls them. Use this for unattended loops (Cowork schedules, CI). Leave it unset if you want every send and publish to wait for your approval in the chat.

Examples

Example prompts

Add 'John Doe <john.doe@example.com>' and 'Jane Doe <jane.doe@example.com>' to the AI newsletter subscribers group

ℹ️ This will:

Create/update the recipients in the database

Add them to the "AI newsletter subscribers" recipient group, creating it if needed

Create a GoodSender template for a weekly AI news digest sent to the AI newsletter subscribers group

ℹ️ This will:

Create a draft email template in GoodSender format following best practices for email template creation

Display interactive preview of this template draft (if AI client supports it)

Increase the number of news in the digest to 5

ℹ️ Done during the template draft creation/editing, this will:

Modify the template accordingly

Display interactive preview of this template draft (if AI client supports it)

Send a test email to me: 'Good Sender <good.sender@example.com>'

ℹ️ Done during the template draft creation/editing, this will:

Send an email with mocked data generated from the current template draft only to you

Save the template as "Weekly AI news"

ℹ️ Done during the template draft creation/editing, this will:

Convert the current template draft to a persistent template stored in the local database

Allow sending emails just by mentioning the template name

Gather this week's AI news and send them using the "Weekly AI news" template

ℹ️ This will:

Create and enqueue a a personalized template-based email for each recipient in the AI newsletter subscribers group

For all recipients who haven't received an email with consent request, will request their consent for receiving emails from you

Send emails to all recipients who granted their email consent

Monitor all recipients with pending consent and send the email as soon as they grant it

Publish a blog post 'Why consent-based emailing wins' and email it to the AI newsletter subscribers group

ℹ️ This will:

Create a draft post from the title and body

Show a signed preview URL

Publish the post

Broadcast it to the group (consent flow still applies)

Telemetry & privacy

Telemetry is enabled by default. MCP server sends aggregate usage and reliability metrics to GoodSender to help improve the product.

Typical signals include:

MCP tool names, call counts, and latency.
Success and error rates.
Short error excerpts on failures.
Background job error codes and frequency.

We do not intentionally collect email bodies, recipient addresses, or template contents as part of this telemetry.

You can opt out of telemetry collection by setting GOODSENDER_TELEMETRY=false in the MCP server environment.

See the GoodSender Privacy Policy for how personal data is handled by the service.