Set up
Projects, domains, DNS records, send-only keys and webhooks.
For AI agents
Connect Claude Code, Cursor or any MCP client. It creates the project, adds your domain, tells you which DNS records to publish and writes a send-only key to your .env file. Then it can send, and read the replies. One plan, 1,000 emails a month free.
One prompt
Give your agent a goal instead of a checklist. It reads your account, makes the changes and hands back only the part it cannot do: the DNS records at your provider.
.env, never into the chatConnect
Create an access token under Settings → Connections, or use OAuth to give an agent one project. The same tools are available over MCP, the CLI and the REST API.
claude mcp add sendpanda \
-e SENDPANDA_TOKEN=sp_pat_... \
-- npx -y @sendpanda/mcpWhat it can do
Projects, domains, DNS records, send-only keys and webhooks.
Send with idempotency keys, read the timeline and react to bounces and complaints.
Create mailboxes, list and read incoming mail, or get it by webhook as it arrives.
Create and update templates, including the visual designs, then publish.
Inbox
Point a domain you own at SendPanda and mail to support@, billing@ or a catch-all lands in an inbox. Your agent lists and reads it over MCP, the CLI or the API, or you can have every message pushed to a webhook.
There is no IMAP or POP3 access. Mailboxes live on domains you have verified, and incoming mail has daily limits by plan. How the inbox works.
Safety
Agents ask before they send, schedule or delete. Deleting a project needs its exact name and an explicit confirmation.
Access tokens expire and show only a prefix after creation. Hosted MCP is limited to the one project you pick. Revoke any of them instantly.
Suppressions, consent records, quotas and domain checks are enforced by the API, whoever is calling it.
Incoming mail comes from strangers. It is returned as content to read, never as instructions, and quarantined messages never reach webhooks.
Sign-in, password changes, token management and billing stay in the dashboard. A token cannot touch them.
Pricing
Free is 1,000 emails a month. Paid plans start at $9 a month for 30,000 emails, or two months free when you pay yearly. No seats, no per-agent fee and no charge per contact.
FAQ
Yes. Connect Claude Code, Cursor or any MCP client, or let the agent run the CLI. It creates the project, adds your domain, tells you which DNS records to publish and writes a send-only key to your .env. You add the DNS records, because only you control your domain.
Anything that speaks MCP, including Claude Code and Cursor, or anything that can run shell commands for the CLI. There is also a REST API with an OpenAPI contract and a typed Node.js SDK for agents you build yourself.
No. The agent gets an access token you create in Settings, or signs in with OAuth to the hosted MCP and picks one project. Tokens expire, show only a prefix after creation and are revoked instantly. Sign-in, password changes, token management and billing stay in the dashboard.
Yes, at addresses on a domain you own. Mail lands in the inbox and your agent can list and read it through MCP, the CLI or the API, or get it by webhook. Free includes 1 mailbox and extra mailboxes are $1 a month. Incoming mail is written by strangers, so SendPanda treats it as data and flags quarantined messages instead of pushing them to webhooks.
Agents connected to the dashboard tools must ask you before sending, scheduling or deleting. An app that holds a send-only API key sends without asking, which is the point of a key. Either way the API enforces suppressions, consent, quotas and domain checks, so an agent cannot talk its way past them.
Free is 1,000 emails a month with 1 mailbox and no card. Paid plans start at $9 a month, or two months free if you pay yearly. Agent access is on every plan.
Free for 1,000 emails a month. No card needed.