Niches · Agent inbox APIs

Send or reply from the agent's inbox as a tool call, once and intact

Two or three companies' docs show a full fit.

2companies, full fit
10companies, fit with a limit
50companies carry the tag
10share of the category's work
7first-hand accounts

What counts as done

The tool schemas load on Anthropic, OpenAI, Gemini and xAI, precisely one message with the intended subject and body is delivered per intent, and an uncertain result can be checked or retried safely.

How agents ask for it: reply to that email, send the update from your inbox, forward this to accounting

Does the job 2

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

ListingWhy
Cherami C
Abderrahmane Gourragui (sole operator, Quebec)
Send requires to, subject and text, and takes an idempotency_key whose receipt reports replayed and idempotency_expires_at, so an uncertain send can be retried safely; drafts send with their own key. source
Postfleet E
Postfleet
The MCP send_email and reply_email tools require client_id, and a repeated client_id replays the first outcome, including rejections and 502s; a 202 pending_approval is documented as success not to retry. source

Does the job, with a limit 10

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
AgentMail API + MCP BB
AgentMail, Inc.
The REST send, reply and draft send take Idempotency-Key, but the MCP send_message and reply_to_message schemas have no key and leave subject and text optional, and the Python SDK retried a keyless send into a double send (a22, open). source
Clawpost E
Frank Harris
send_email requires to, subject and body, so blank mail can't go out, but sends carry no idempotency key and the tools have no annotations, so an uncertain send can't be retried safely. source
Hostinger Agentic Mail C
Hostinger International Limited
POST /send takes to, subject, text, inReplyTo and forwardOf, but no field is required in the schema and there is no idempotency key, so blank mail or a double send after a timeout is possible. source
Inbound C
EXON ENTERPRISE LLC
Send and reply take Idempotency-Key and a mailbox credential can only send as its own identity, but the MCP tool schemas weren't dumped against provider rules. source
InboxAPI D
Sitka Capital Pty Ltd
The Codex schema rejection (a23) was fixed in May 2026 and new recipients need allow_new_recipients, but sends take no idempotency key and the hosted tools' input schemas aren't published. source
Lumbox B
Cyphen Innovations Private Limited
Sends and replies honour Idempotency-Key, but the hosted MCP's tool schemas weren't dumped against provider rules, and the MCP package was last released in June 2026. source
Mailows C
DW Technology LLC
Every send carries a stable Idempotency-Key honoured for 24 hours and the allowlist rejects wrong recipients, but there is no reply operation, and the MCP schemas weren't dumped. source
mails.ai Agent Email D
Mails.ai
mails_send and the reply tool take idempotency_key, and tool names use underscores for Claude and OpenAI rules, but mails_send requires only to and subject, so a body can be empty, and a call without a key gets a fresh one, so a model's retry can still send twice. source
MailSlurp B
Pettman OÜ
Reply and draft send take Idempotency-Key, but the hosted MCP's tool names use dots (mailslurp.reply_to_email), which the Anthropic and OpenAI tool-name pattern rejects unless the client renames them. source
Robotomail B
Tiny Bot Labs Limited
Send requires subject and bodyText, so blank mail is refused, but there is no idempotency key, so an uncertain send must be checked by listing outbound messages before a retry. source

What stops agents finishing

  • schema rejected by the model API
  • optional subject and body let a model send blank mail
  • SDK retries a send with no idempotency key
  • OAuth refresh drops the MCP session

Around it

Before: agent-inboxes.multi-day-thread. After: agent-inboxes.multi-day-thread.

Where this comes from

Weight 10 of 100 in Agent inbox APIs: 7 of 31 accounts are integration failures on send (a17, a18, a19, a21, a22, a23, a24); weighted at 10 rather than by count because they describe how the other jobs fail, not a separate goal

The same as data: /api/v1/use-cases/agent-inboxes.reply-as-tool.json, or the MCP tool get_use_case with id agent-inboxes.reply-as-tool. 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.