Aldus + Model Context Protocol (MCP): Drive your newsletter from Claude, ChatGPT, Cursor, or any AI agent
Aldus is an MCP-native newsletter platform — every core operation is exposed as a Model Context Protocol server at usealdus.com/api/mcp. AI agents in Claude, ChatGPT, Cursor, Windsurf, and any MCP-compatible client can plan, draft, schedule, send, and analyse newsletter issues programmatically. No glue code, no custom plugins.
Start free →What You Can Do with Model Context Protocol (MCP) + Aldus
- Run an entire weekly newsletter from a Claude conversation — agent drafts, you approve in the same chat
- Use Cursor to scaffold a newsletter for a new project alongside the codebase
- Build an autonomous content agent that reads a Notion source, drafts an issue in Aldus, and queues it for approval — all in a single agent loop
- Run multi-newsletter agencies via Claude with one MCP connection — switch newsletters per command
- Generate analytics reports on demand by asking the agent for engagement breakdowns
How the Integration Works
Generate an Aldus API key
Create a scoped API key in Settings → API. Keys are prefixed aldus_sk_ and stored as a SHA-256 hash on our side.
Add Aldus to your MCP client config
Point your client at https://usealdus.com/api/mcp with a Bearer auth header. Works with Claude Desktop, Cursor, Windsurf, and any MCP-compatible client. Local stdio transport is also available for development.
The agent discovers 14 tools automatically
Tools cover newsletters, issues, subscribers, sending, and analytics. The agent can plan multi-step flows like 'draft an issue, send a test, approve, send to all subscribers' as a single chain.
Run, observe, iterate
Every tool call is rate-limited and metered against your plan. Sessions are tracked via the mcp-session-id header so the agent maintains context across calls.
Why Model Context Protocol (MCP) Works Better with Aldus
Set Up in Under 5 Minutes
- Generate an API key in your Aldus dashboard at Settings → API
- Add the MCP server config to your client (see snippet below)
- Ask your agent: 'List my Aldus newsletters' — confirm the connection works
- Try a real workflow: 'Draft this week's issue of [newsletter name] from these three sources' — the agent calls create-issue and returns the draft for approval
Frequently Asked Questions
Which MCP clients are supported?
Aldus exposes a standard Streamable HTTP MCP server, so any compliant client works — Claude Desktop, Cursor, Windsurf, Cline, Continue, and custom agents built on the official MCP SDKs (TypeScript, Python, Rust, Swift). For Claude Desktop on macOS, add the server to the mcpServers section of claude_desktop_config.json with the URL and Bearer header.
How is the MCP server authenticated?
Per-account API keys, prefixed aldus_sk_. Send as Authorization: Bearer aldus_sk_yourkey. Keys can be scoped (read-only, send-only, full) so a research agent gets analytics access without the ability to send to your list. Sessions are tracked via the mcp-session-id header returned on initialise.
What can the agent actually do?
All 14 tools: list-newsletters, get-newsletter, create-newsletter, create-issue (triggers AI drafting), get-issue, approve-issue, send-issue, send-test-email, list-subscribers, add-subscriber, import-subscribers (up to 500 at a time), get-analytics, get-issue-analytics, get-account-status. The agent can chain these into multi-step workflows — for example, draft → test-send → approve → send-to-all is one conversation.
Is there a stdio transport for local development?
Yes. Install @workspace/aldus-mcp and run the stdio entry point with your API key as ALDUS_API_KEY in the env. This is useful for testing agent loops locally before pointing them at the production HTTP endpoint.
Are MCP tool calls rate-limited?
Yes. Calls count against your plan's API quota — same metering as direct REST API calls. Free tier gets 1,000 calls per month, Starter 10,000, Growth 50,000, Pro unlimited. send-issue is metered separately against your monthly send allowance.