Niches · Agent inbox APIs

Act on mail forwarded or sent to the agent (invoices, listings, requests)

Two or three companies' docs show a full fit.

3companies, full fit
9companies, fit with a limit
31companies carry the tag
8share of the category's work
3first-hand accounts

What counts as done

Each message triggers the agent once (webhook or event, deduplicated), with text, sender and attachments readable, and the outcome per message recorded.

How agents ask for it: process whatever gets emailed to this address, forward receipts here and file them, read the job alerts and shortlist them

Does the job 3

The docs show it meets the test above, and nothing we read says otherwise.

ListingWhy
AgentMail API + MCP BB
AgentMail, Inc.
Webhooks (signed, at-least-once), WebSocket events and polling webhooks deliver each message, extracted_text and attachment text extraction make bodies readable, and receive allow lists and Agent Armor screen senders; a04 processes job alerts and emailed documents this way. source
mails.ai Agent Email D
Mails.ai
Each message.received event carries injection_score and sender_reputation, webhooks are HMAC-signed, retried with backoff, logged and replayable, and events can be redelivered, so each message triggers the agent with a recorded outcome. source
Postfleet E
Postfleet
Each message.received webhook is signed, and every message carries comprehension_status (injection screening), a sanitisation report and optional schema extraction, with dead-lettered deliveries redriven from the dashboard. source

Does the job, with a limit 9

It does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.

ListingWhy
Cherami C
Abderrahmane Gourragui (sole operator, Quebec)
Signed webhooks with delivery attempts and event listing can trigger the agent, and messages carry text, HTML and attachments, but there is no sender allowlist or injection verdict, and without a webhook mail just waits. source
Clawpost E
Frank Harris
Mail is stored with attachments and searchable, and unknown senders are held for approval by default, which the map names as a blocker. Nothing triggers the agent per message. source
Hostinger Agentic Mail C
Hostinger International Limited
Webhooks per mailbox and event, with a test button and delivery log, trigger the agent, and the text and attachment endpoints read the message. There is no receive allowlist or injection verdict, and failing webhooks are paused. source
Inbound C
EXON ENTERPRISE LLC
Webhooks carry parsed text and attachment download URLs, and Guard rules allow, block, junk or route mail by sender or by a natural-language prompt. The webhook carries a static verification token, not a signature; the last assessment found failed deliveries aren't retried. source
InboxAPI D
Sitka Capital Pty Ltd
Every message carries trust_level and trust_reason, and untrusted bodies are datamarked (spotlighting), but nothing triggers the agent per message; it must poll and track what it has handled. source
Lumbox B
Cyphen Innovations Private Limited
Webhooks (from Starter) and the SSE stream trigger the agent per message, with parsed JSON, codes and links, but there is no sender allowlist or injection verdict, and webhooks aren't on Free. source
Mailows C
DW Technology LLC
A per-mailbox webhook with a rotating secret triggers the agent and subject and body come only inside untrusted_content, but reading marks mail read with no undo, so the agent must keep its own processed-id record. source
MailSlurp B
Pettman OÜ
Webhooks with reliable-processing guidance, routing rulesets and attachment text extraction cover forwarded mail, but no injection verdict is attached; the docs only tell agents to treat bodies as untrusted. source
Robotomail B
Tiny Bot Labs Limited
Webhooks (at-least-once, deduplicated on X-Robotomail-Delivery-Id) and SSE trigger the agent per message, but Free discards mail from anyone but the owner, and there is no allowlist or injection verdict. source

What stops agents finishing

  • unknown senders held until a person approves (Clawpost)
  • free plan discards mail from anyone but the owner (Robotomail)
  • no per-message outcome after a restart

Around it

Before: agent-inboxes.own-identity. After: document-extraction.parse-invoice, accounting.record-bill.

Where this comes from

Weight 8 of 100 in Agent inbox APIs: 3 of 31 accounts: job-listing and bookkeeping mail (a04), webhook per arriving message (a09), and the request for unsolicited inbound to an agent-owned address (a28)

The same as data: /api/v1/use-cases/agent-inboxes.inbound-forward-processing.json, or the MCP tool get_use_case with id agent-inboxes.inbound-forward-processing. Other jobs in this category: /api/v1/niches?category=agent-inboxes.

Draft, read 2026-11-15 from public docs and first-hand accounts, not yet reviewed by a person. A mark says what the docs show, not what a run measured.

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.