Skip to content
RRAECHLY

Model Context Protocol · Streamable HTTP

The whole product, as tools.

Raechly exposes all 22 of its capabilities over MCP. Not a reporting subset. Your assistant registers sending domains, reads back DNS records, verifies identities, warms mailboxes, imports leads, writes sequences, fires the pipeline and reads the send log with the errors intact.

Connect
claude mcp add --transport http raechly \
  https://reachly-mcp-server.capaxe.workers.dev/mcp \
  --header "Authorization: Bearer rk_your_key"

No key yet? Add the server without the header and ask your assistant to run register_account. It is the one tool an anonymous connection can see.

A full setup, in one message

The reason all 22 tools matter is that setup is the hard part, and setup is exactly what most integrations leave out.

You say

Set up mail.acme.com for outbound in eu-west-1, create two mailboxes, start warmup on both, then draft a three-step sequence for agency owners and leave it in draft.

It runs

  1. 1. create_domain
  2. 2. get_domain_dns → shows you the records
  3. 3. check_domain → until verified
  4. 4. create_mailbox ×2
  5. 5. start_warmup ×2
  6. 6. create_warmup_pool_account
  7. 7. create_campaign → draft

All 22 tools

Account

Domains

Mailboxes

Warmup

Leads

Campaigns

Pipeline

Why the setup tools matter more than the reporting tools

Several cold email platforms now publish MCP servers, and most of them are good at the same thing: reading campaign stats and managing leads. That is genuinely useful, and it is also the easy half.

The half that costs people weeks is infrastructure. Registering a sending identity, publishing DKIM and MAIL FROM and DMARC records correctly, confirming verification actually completed, provisioning reply capture, and then working out which of those steps is the reason nothing is sending. Raechly exposes all of it, which means the assistant that helps you write the sequence is the same one that tells you your third DKIM CNAME is missing.

If you are comparing MCP surfaces, count what you can accomplish end to end rather than the number of tools. Ours is 22 because the product is deliberately narrow. See how that stacks up against Smartlead and Instantly, both of which ship their own.

MCP questions

What is an MCP server for cold email?

The Model Context Protocol is an open standard from Anthropic that lets an AI assistant call external tools. A cold email MCP server exposes your sending platform as tools the assistant can invoke, so you describe what you want in plain language instead of clicking through an interface.

Which clients can connect to the Raechly MCP server?

Anything that speaks Streamable HTTP MCP. That includes Claude, Claude Code, Cursor and VS Code. Add the endpoint, set the Authorization header to your organization API key, and the tools appear.

Is my API key sent to Anthropic?

No. The key is held by your MCP client and sent to the Raechly MCP server as a bearer token, which forwards it to the Raechly API. It is the tenant boundary: the API resolves your organization from the key.

What can the assistant do without a key?

Only register_account. An anonymous connection sees a single tool, which creates an account and returns the key you then reconnect with. Everything else requires authentication.

Can the assistant send email without me approving it?

Tool calls follow your client’s approval settings, and the destructive tools are annotated so clients can prompt before running them. Beyond that, the same server-side guardrails apply to every path: send windows, daily caps and the reply-capture gate do not care whether a human or an assistant made the call.

Add the server, ask it to register you.

Anonymous connections see one tool. Run it, save the key, reconnect, and the other 21 appear.