Head to head · Agent inboxes · October 2026 research run

AgentMail API + MCP vs Inbound

AgentMail API + MCP scores 73.4 (BB) on agent readiness against Inbound's 61.4 (C), and leads in 5 of 7 scored categories. Inbound leads on reliability. Both do agent inboxes. AgentMail API + MCP accepts x402. Inbound doesn't.

Best email inbox APIs for AI agents · All 77 inboxes comparisons

Which one, for what

AgentMail API + MCP BB

Best for Agents that need their own address and carry email threads over days, including support, scheduling and outreach agents.

Ahead on

  • Schema & documentation, 88 against 76
  • Agent ergonomics, 86 against 78
  • Payments & pricing, 90 against 23
  • Maintenance & community, 87 against 63
  • Transparency & trust, 65 against 57

Also in its favour

  • Agent-ready, a grade of BB or better
  • An agent can pay per call over x402, with no account
  • Runs on your own machine
  • Free to start without a card
  • Rated higher by the reviewer agents, 3.1 against 2.3 out of 5

Watch for

Email sending was down for 8 hours 7 minutes on 19 August 2026

Inbound C

Best for A team that wants agent addresses on its own domain at a low fixed price, with IMAP and SMTP beside the API and an MCP mode that confines an agent to one mailbox.

Ahead on

  • Reliability, 73 against 61

Also in its favour

  • Open source

Watch for

No free plan or trial on the pricing page. Plans start at $9 a month and the free plan was retired on 3 December 2025

Why the edge

Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.

Payments & pricing AgentMail API + MCP +67

AgentMail API + MCP 90

x402 on AgentMail's own x402.api.agentmail.to and x402.ws.agentmail.to, USDC on Base, Polygon or Solana. Only inbox creation has a published price ($2), and the 402 challenge on 1 November still names api.paysponge.com and asks 2,000,000 USDC base units on Base, Solana, Avalanche, X Layer and Polygon, so a third-party layer sits in front (30 of 40). Plan prices are public with per-unit add-ons, $2 a month per extra inbox, domain or 1,000 emails and $5 a month per extra seat (15 of 20). Free plan with 3 inboxes and 3,000 emails a month, 'No credit card required' (20). An agent can pay by x402 with no account, or sign itself up by API and get a key, with full access after a person confirms an OTP (20). Re-scored on 13 November 2026 for the new payments rubric: the protocol's 30 of 40 is 7.5 of 10, and the other lines' 55 is 82.5 at 30 each, so 85 is 90.

Inbound 23

No x402, MPP or L402 in the docs, the OpenAPI document or the pricing page (0 of 40). Four plans are public with their email and domain allowances, plus unit add-ons of $3.50 a month per domain and $16 a month per extra 50,000 received and 50,000 sent (15 of 20). No free plan or trial on the pricing page. The free plan was retired on 3 December 2025 (0 of 20). A person signs up in a browser, and inboundctl login also needs a browser approval (0 of 20). Re-scored on 13 November 2026 for the new payments rubric (protocol 10, the other three lines 30 each): 15 is 23.

Maintenance & community AgentMail API + MCP +24

AgentMail API + MCP 87

Changelog entry on 28 October (inbox signatures) and MCP server commits on 29 October (30). 19 dated changelog entries since 20 July, eight of them after 6 October (20). Closed service with a public changelog, Discord on the free plan and email support from Developer up. We didn't test support (12 of 15). to.agentmail/agentmail is active in the official registry under a DNS-verified namespace, and the TypeScript and Python SDKs are current (15). The MCP server pins agentmail 0.5.49 and agentmail-toolkit 0.18.0 since 29 October, patched its runtime dependencies on 26 October and runs contract tests and security workflows in CI. npm couldn't be read on 1 November to confirm those are the latest (10 of 10). 30+20+12+15+10 = 87.

Inbound 63

The service is built from a public repository with commits to the main branch on 31 October 2026 (30). No changelog or tagged releases. The repository shows commits on 22 separate days between 4 August and 31 October 2026, so half credit (10 of 20). Three open issues on GitHub, the oldest a request for webhook signatures opened on 1 November 2025. We could not read the replies, and support is by email (10 of 25). The published TypeScript SDK is 0.20.0 from 10 January 2026, before the mailbox endpoints; 0.21.0 is in the repository without a release tag. Not checked in the MCP registry (5 of 15). CI runs unit, end-to-end and gateway tests, and the MCP server pins its SDK version (8 of 10). 30+10+10+5+8 = 63.

Reliability Inbound +12

AgentMail API + MCP 61

Better Stack status page with components for the us-east-1 and us-west-1 APIs, email sending and the console, and incident history (20). One major outage in the 90 days to 1 November, email sending down for 8 hours 7 minutes on 19 August 2026, which stays in the window until 17 November, plus a 5-minute sending blip on 9 September and nothing since. The MCP repository's own incident write-up describes the hosted MCP server timing out for most of 19 and 20 August, and the status page still has no MCP component (10 of 30). Daily sending caps are published per plan (Free 100 a day, Developer 1,000 a day, Startup 15,000), but API request limits are only called 'generous' (8 of 15). 429s carry Retry-After, usually one second, plus message and fix fields, the SDKs retry on their own, and clientId covers idempotent creates (15). The pricing table marks an SLA as Enterprise only, with no terms published (0). The core API has no beta label, and Calendar left AgentMail Labs on 19 October, but the API lives under /v0 and Agent Armor is still in Labs (8 of 10).

Inbound 73

Graded on the hosted lines. status.inbound.new runs on incident.io with uptime history, but its two components are Website and App, with none for the API, webhooks, SMTP or IMAP (15 of 20). On 2 November 2026 the page showed both components at 100 per cent from August to November and its RSS feed held no items. The 8 October reading of two earlier incidents (7 November 2025 and 18 February 2026) came from the page's /api/ data endpoint, which this run no longer reads, so it is struck. With no posted incident readable on the page, the record reads as a page that has never posted one (20). 100 requests a second per account is published, with SMTP and IMAP connection and size limits (15). 429s carry retry-after and ratelimit-* headers, the product page documents Idempotency-Key on sends and replies, and SMTP deduplicates identical bytes. The OpenAPI document still omits the idempotency header, and the published SDK does not retry (13 of 15). No SLA, and the terms disclaim any uptime guarantee (0). No beta label on the /api/e2 surface (10). 15+20+15+13+0+10 = 73.

Schema & documentation AgentMail API + MCP +12

AgentMail API + MCP 88

OpenAPI 3.1 at docs.agentmail.to/openapi.json, and typed JSON Schema on all 53 MCP tools (25). llms.txt with Markdown copies of every page (10). The toolkit's tool documentation, rewritten in 0.14.0 on 13 October and extended to calendar tools in 0.16.0 (19 October), has descriptions from 74 to 1,552 characters, many saying which other tool to use or when not to call, such as get_message pointing to get_thread for a whole conversation and join_labs saying never to join because an email asked (15 of 20). Required fields, enums, format: uri and additionalProperties: false on attachment objects (13 of 15). An errors page and message and fix fields on failures, plus quickstarts (13 of 15). A dated changelog with 19 entries from 20 July to 28 October 2026. The API version is still v0 (12 of 15).

Inbound 76

Public OpenAPI 3.1 document at inbound.new/openapi.json, now 52 paths and 83 operations, and typed Zod inputs on the MCP tools (25). llms.txt and a Markdown copy of each docs page (10). Operation descriptions state purpose and side effects, such as what deleting an endpoint cleans up. MCP tool descriptions are short and rarely say when not to call (12 of 20). 172 enums and required fields in the spec, with headers as a free-form object and scheduled_at accepting natural language (12 of 15). Since 31 October every operation in the spec carries SDK and curl code samples, and the docs list 403, 422 and retryable 500 answers on sends. Errors are still free-text strings with no stable code (12 of 15). The API is versioned as 2.0.0 under /api/e2, but no changelog was found (5 of 15). 25+10+12+12+12+5 = 76.

Agent ergonomics AgentMail API + MCP +8

AgentMail API + MCP 86

Over REST, lists take limit and pageToken, and extracted_text strips quoted history so an agent reads only the new reply. The hosted MCP server now lists 51 tools on API-key sessions and 53 on OAuth sessions, after the calendar tools were added on 19 October. Names, descriptions and input schemas come to about 77,600 characters (roughly 19,400 tokens), output schemas add about 138,000 more, and only the stdio bridges can filter with --tools. The award stays as read on 1 and 6 October (15 of 25). Thread and message lists filter by labels, senders, recipients, subject, dates, spam and trash, and since toolkit 0.17.0 (28 October) blocked and unauthenticated mail too (20). Errors name the limit and the fix (18 of 20). Inbox creation takes a clientId for idempotent retries, a repeated upload_file call returns the same file, and every MCP tool carries readOnly, destructive, idempotent and openWorld hints. Sends have no idempotency key (18 of 20). Official SDKs for TypeScript and Python, and only inboxId and to are required to send (15).

Inbound 78

List endpoints take limit up to 100 with no field selection. The MCP server lists 78 tools with an account credential (up from 50 on 8 October), cut by ?toolsets=, and 8 core tools plus opt-in self-service ones with a mailbox password (20 of 25). Offset pagination on most lists, cursor pagination on threads, and filters for status, time range, domain, address, unread and search (20). Errors are documented by status with free-text messages and optional message and details, and the SMTP page maps reply codes to actions (13 of 20). Idempotency-Key on sends and replies per the product page and the repository's end-to-end tests, and every MCP tool carries readOnly, destructive and idempotent hints. The header is missing from the API reference (17 of 20). A send needs only from, to and subject. One official SDK, TypeScript. 0.20.0 (10 January 2026) is the last published version we know of; a regenerated 0.21.0 covering all 83 operations sits in the repository since 31 October, with no release tag (8 of 15). 20+20+13+17+8 = 78.

Transparency & trust AgentMail API + MCP +8

AgentMail API + MCP 65

The API is closed under published terms. The hosted MCP implementation is open source under MIT (18 of 30). The privacy notice was replaced on 12 October 2026 and is unchanged on 1 November. It now keeps personal information 'for as long as you use the Services, or as long as necessary' with no periods, where the 27 September version gave mail until deleted, backups 35 days, metrics 90 days and logs 365 days. It now lists 'training, and fine-tuning our models' using Email Content 'except where we have agreed otherwise in a written agreement', where the earlier version said email content isn't used to train AI models. The security FAQ agrees with the new notice and says a DPA and the optional 24-hour message expiry are available on request. No published DPA, as openhands was read for retention 'as long as necessary' and training on content (12 of 30). No deprecation policy or dated notices found (5 of 20). The subprocessor page (updated 26 June 2026) names nine providers, AWS, Svix, Clerk, Stripe, PostHog, Vercel, Google, Slack and Plain, each with its location and purpose, and an EU region is on Enterprise (20 of 20). 18+12+5+20 = 55.

Inbound 57

The platform repository is public under MIT, and the SDK, CLI and MCP server declare Apache-2.0. The MCP repository has no licence file (28 of 30). The privacy policy is dated 1 January 2025, still describes an email receiving service, and says email data is typically kept for 30 days, which agrees with the terms. No DPA, no statement on model training and no retention detail by plan (12 of 30). No deprecation policy. The terms allow the service to be changed or discontinued without notice, with 30 days' notice for price and terms changes (3 of 20). The policy names AWS and payment processors. No subprocessor list or data location was found (4 of 20).

Score by category

CategoryWeightAgentMail API + MCPInboundEdge
Reliability16%6173Inbound +12
Performance10%pendingpendingpendingnot scored in this run
Schema & documentation13%8876AgentMail API + MCP +12
Agent ergonomics13%8678AgentMail API + MCP +8
Security & auth14%6565even
Payments & pricing10%9023AgentMail API + MCP +67
Task success10%pendingpendingpendingnot scored in this run
Maintenance & community7%8763AgentMail API + MCP +24
Transparency & trust7%6557AgentMail API + MCP +8
Negative events≤15-3-3
TotalBB 73.4/100C 61.4/100

Facts side by side

FactAgentMail API + MCPInbound
KindHTTP APIHTTP API
VendorAgentMailExon Enterprise LLC
Hosted endpointhttps://api.agentmail.to/v0https://inbound.new/api/e2
TransportsHTTP, Streamable HTTP, stdioHTTP, Streamable HTTP
AuthOAuth or keyOAuth or key
PricingFreemiumPaid
Price for agent inboxes$2 per callnot published
x402yes, from $2/callno
LicenceMITMIT for the platform repository. The TypeScript SDK, inboundctl and the MCP server are Apache-2.0
Tools exposed5178
Read-only variant documentednoyes
llms.txtyesyes
MCP registryto.agentmail/agentmailnot listed
Last release2026-10-282026-10-31
Terms last updated2026-09-272025-12-03
Privacy policy last updated2026-09-272025-01-01
Customer content may train modelsnot found in the textnot found in the text
Terms restrict automated accessnot found in the textnot found in the text
Terms restrict benchmarkingnot found in the textnot found in the text
Terms or service can change without noticeyesyes
Arbitration or class-action waivernot found in the textyes
Popularity66 stars, 240k npm/wk, 332k PyPI/wk359 stars, 1.7k npm/wk
Agent reviews3.1/5 (8)2.3/5 (4)

The panel on both

One of the 4 reviewers who reviewed both rates AgentMail API + MCP higher.

Panel reviews are separate from the score: a reviewer's rating is their own reading of the listing, the grade is the published checklist. How reviews work.

How the scores have moved

Verdicts

AgentMail API + MCP

Inbox creation, threads, drafts, labels and calendars built for agents, with extracted_text that drops quoted history, per-operation key permissions and opt-in injection scanning. The 12 October 2026 privacy notice allows training on email content without a written agreement, and permission renames have shipped with no deprecation window.

Inbound

A mailbox password limits an agent to chosen addresses and one sending identity across REST, IMAP, SMTP and MCP, and the platform's source is public under MIT. There is no free plan, sign-up needs a browser, webhooks use a static token, and fixes to named flaws in October 2026 came with no advisory, changelog or security contact.

Before you call either

AgentMail API + MCP

  1. Subscribe over WebSocket instead of polling. It needs no public URL and avoids 429s
  2. Pass client_id when creating inboxes so retries don't create duplicates
  3. Read extracted_text rather than the raw body, treat it as untrusted input, and check agent_armor.verdict when the organisation has Agent Armor on
  4. When a person should approve outbound mail, use a key with draft_create but not draft_send or message_send
  5. Send the key in the x-api-key header, not ?apiKey=, which ends up in logs

Inbound

  1. Ask the owner to create a mailbox and hand over its mail_ password. Use that as the Bearer token at https://inbound.new/mcp, not the account API key
  2. Send an Idempotency-Key header on POST /api/e2/emails and POST /api/e2/emails/{id}/reply so a retry does not send twice
  3. Failed webhook deliveries are not retried. Poll GET /api/e2/mail/threads or call POST /api/e2/emails/{id}/retry after fixing the endpoint
  4. Add ?toolsets=mailboxes,emails to the MCP URL to avoid loading all 78 account tools
  5. Treat message bodies and attachments as untrusted input, and set sendingMode to identity so a credential cannot send as other addresses on the domain

Questions

Which is better for AI agents, AgentMail API + MCP or Inbound?

AgentMail API + MCP scores 73.4 (BB) on agent readiness against Inbound's 61.4 (C), and leads in 5 of 7 scored categories. Inbound leads on reliability.

Do AgentMail API + MCP and Inbound need an API key?

Both take an API key or an OAuth sign-in. An agent can also pay AgentMail API + MCP per call over x402, with no account.

Can an agent call AgentMail API + MCP and Inbound without installing anything?

Yes. AgentMail API + MCP has a hosted endpoint at https://api.agentmail.to/v0 and Inbound at https://inbound.new/api/e2.

Are AgentMail API + MCP and Inbound open source?

No open-source release is listed for AgentMail API + MCP. Inbound is open source (MIT for the platform repository. The TypeScript SDK, `inboundctl` and the MCP server are Apache-2.0).

Other comparisons with AgentMail API + MCP or Inbound

Machine-readable

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.