Receive on any address
Every local part on the project’s inbound domain works immediately. Use one address per agent, customer, ticket, or run without creating inbox objects.
Receive, read, reply, forward, and send through one API. No Gmail OAuth, no inbox provisioning, and no polling loop—just an address your software owns.
View as MarkdownEach project gets a catch-all inbound domain. There is no inbox object to create: support@, agent-42@, and invoice-918@already work. Choose an address scheme that matches your application state.
The address routes the message; it does not grant authority. Your application still resolves which agent owns the conversation, which tools it may use, and which external actions require a person to approve them.
Read the email inbox API guideThe API carries the message identity and context an agent needs without turning every incoming sentence into an instruction.
Every local part on the project’s inbound domain works immediately. Use one address per agent, customer, ticket, or run without creating inbox objects.
Use a signed email.received webhook for a backend or a WebSocket stream for a long-running agent. Fetch the full message only when it matters.
Retrieve text and HTML bodies, participants, raw threading headers, authentication results, spam verdicts, and the attachment list over API.
Reply against the inbound message ID. Email Bump sets In-Reply-To, References, Reply-To, and the subject so the conversation stays together.
Inspect metadata before downloading bytes. Keep untrusted files outside privileged execution and hand only the required content to the agent.
Forward one message or create a standing rule. The original formatting and attachments survive, and Reply-To points back to the sender.
Push a summary first, fetch sensitive content only when required, and keep every consequential action behind application policy.
signed email.received event
policy, tools, and approval
threaded reply from verified domain
The reply endpoint derives conversation headers from the message you are answering. Your agent supplies the response, not a hand-built MIME thread.
# Find messages waiting for the agent
curl 'https://emailbump.com/api/v1/inbound?limit=10' \
-H "Authorization: Bearer $EMAILBUMP_API_KEY"
# Read the selected message
curl 'https://emailbump.com/api/v1/inbound/MESSAGE_ID' \
-H "Authorization: Bearer $EMAILBUMP_API_KEY"
# Reply in the original conversation
curl -X POST \
'https://emailbump.com/api/v1/inbound/MESSAGE_ID/reply' \
-H "Authorization: Bearer $EMAILBUMP_API_KEY" \
-H 'Content-Type: application/json' \
-d '{ "text": "It ships tomorrow." }'Use addresses as durable routing primitives for conversations your software is actually responsible for.
Give each queue or customer a stable address. Classify, draft, answer, or route the conversation to a human.
Let an agent sign up for a service, receive the confirmation message, and return the code to the authorized workflow.
Receive invoices and statements at task-specific addresses, inspect attachments, extract approved fields, and retain the original.
Turn replies, documents, approvals, and exception messages into durable events for your application.
Treat email bodies and attachments as user-controlled data, never system instructions.
Use SPF, DKIM, DMARC, spam, and virus results as identity and risk signals—not permission.
Give the agent narrow tools and scope each credential to the project it actually needs.
Require a human confirmation before forwarding sensitive data or sending consequential replies.
Record message IDs, decisions, tool calls, and outbound outcomes for later review.
Every operation is available as a direct API. The CLI, MCP server, and open agent skills put safer workflows around the same account.
It is a programmatic surface that lets an agent receive, inspect, send, reply to, or route email without operating a traditional mail client. A useful agent API exposes message identity, threading, attachments, delivery events, and constrained actions—not only a generic send call.
No. Email Bump gives a project a catch-all inbound domain, so every local part already receives. You can assign addresses per agent or task without provisioning mailbox records, while application state remains the source of authorization and ownership.
Yes. Reply using the received message ID and a verified sending identity. Email Bump preserves the thread headers and points Reply-To back to the inbound address so the next response returns to the same workflow.
No. SPF, DKIM, and DMARC help establish which domain sent a message; they do not make the message’s instructions trustworthy. Treat bodies and attachments as untrusted input, constrain tools, and require approval before consequential sends, forwards, purchases, or account changes.
Not when the agent needs an address of its own. Gmail API and Microsoft Graph are appropriate when a user wants an application to act inside an existing mailbox. An inbound email API is simpler and more private for new application-owned addresses.
Yes. The same Email Bump project can send transactional messages, receive replies, manage contacts and consent, run campaigns and flows, and expose the CLI, MCP, and agent-skill surfaces.
Start free with inbound email, transactional sending, API, CLI, MCP, campaigns, and flows in one project.