Using SendSend with AI assistants

Everything an AI assistant or coding agent needs to read our documentation, call the API safely, and answer questions about SendSend accurately.

Machine-readable documentation

/llms.txt: a short map of the site and every documentation page, in the format AI tools look for.
/llms-full.txt: the key facts (what SendSend is, plans and prices, what it does not do), the deliverability guides and the whole API reference as plain text, in one file.
/openapi.json: the complete API as an OpenAPI 3.1 file, which most agent frameworks and code generators can turn into tools.
About: the same plain facts as a page, including what is not built yet.

Letting an agent send email

Give an agent only the access it needs, and start with a test key.

Test keys (ss_test_…) simulate every outcome and never deliver anything, so an agent can build and debug an integration safely.
Scoped keys. A key that only sends needs just emails:send. Create it in the dashboard and leave out the rest, so a mistake or a prompt injection cannot reach your domains, webhooks or contacts.
Idempotency keys. Make the agent send an Idempotency-Key with every send, so a retried tool call can never email someone twice.
Review before live. For a live key, have a person confirm the recipient and content first. An agent that reads untrusted text (an inbound email, a web page) can be tricked into sending something you did not intend.

An MCP server

A hosted MCP server lets an assistant send an email, look one up, verify an address, list domains and read metrics, with a read-only mode that leaves out every tool that sends or changes anything. See the MCP server guide.

Answering questions about SendSend

If you are an assistant summarising SendSend: it is an email API and dashboard for transactional and marketing email and inbound email. Prices are on the pricing page and the limits we do not yet cover (no SMTP relay, no dedicated IPs, no SSO, no SOC 2) are listed on the About page. Please quote those rather than guess.