How to give an AI agent its own inbox for verification codes and sign-in links

Many sites ask "check your email" at sign-in. They send a six-digit code or a one-time link. If the code goes to your personal inbox, the agent stops and waits for you. If you forward your mail to the agent, it reads everything else in that inbox too. A separate address on the agent's own seat keeps its codes in one place, apart from your own mail and your own login.

Here are common options. Every entry for another tool comes from that tool's own page. Not stated means the tool's own pages did not say, as of October 24, 2026.

• ›What it gives the agent: One inbox address per identity, together with saved logins, authenticator codes, saved sessions, site rules, approvals and a log
• ›How the code reaches the agent: runnerID catches the code and types it into the page, or opens the sign-in link
• ›Free tier or price: Free tier with 1 identity and 100 logins and codes a month; Solo $12/mo (1 identity), Builder $29/mo (5), Team $99/mo (25)

• ›What it gives the agent: Inboxes created through a REST API, with send, receive and threads; an MCP server; Python and TypeScript SDKs [1]
• ›How the code reaches the agent: Webhooks and WebSockets deliver new mail to your code [1]
• ›Free tier or price: Free tier with 3 inboxes and 3,000 emails a month [1]

• ›What it gives the agent: An agent identity that owns one mailbox, with optional phone number and vault access [2][3]
• ›How the code reaches the agent: Your code reads the inbox through the SDK, CLI or IMAP [3]
• ›Free tier or price: Not stated

• ›What it gives the agent: An inbox per agent, plus phone calls, a secrets vault, public endpoints, permissions and an audit log [4]
• ›How the code reaches the agent: Not stated
• ›Free tier or price: Free plan; Builder $19/mo; Team $149/mo [4]

• ›What it gives the agent: A dedicated inbox per agent on your domain [5]
• ›How the code reaches the agent: Webhook or WebSocket with the parsed message as JSON; your agent extracts the code [5]
• ›Free tier or price: Free tier with 3 inboxes and 3,000 emails a month [5]

• ›What it gives the agent: A short-lived inbox bought per job [6]
• ›How the code reaches the agent: An endpoint returns the extracted code or link [6]
• ›Free tier or price: $0.05 by x402 or $0.50 by card for a one-hour inbox [6]

The main difference is what happens after the email lands. An inbox API gives the message, or the code, to your code or your agent. runnerID can type the code into the page itself: over MCP, fill_email_code types it and does not return it, so the model does not see it. For hosted agents whose browser runnerID cannot reach, get_email_code returns the code to the agent, and each release is logged [7].

• ›1. In the owner console at https://api.runnerid.torchrunner.ai/console, open the identity. Its inbox address is shown there.
• ›2. Use that address as the email on the agent's own seat (a staff or delegate login the site offers), or on an account the site lets you share. Don't move your own account's email to it.
• ›3. When the agent reaches the "enter the code" step, it calls fill_email_code with the site and its browser. Over a helper, the call is emailCode(site) or email_code.
• ›4. runnerID waits for the email, takes the code, and types it into the page. The model never receives it.

Each identity also holds the site's saved login and authenticator key, so one call sequence can cover the password, the authenticator code and the emailed code.

Some sites send a link instead of a code. runnerID's open_magic_link tool waits for the link and opens it only when the link's host is the site the agent is signing in to, or a subdomain you marked as trusted. If the host does not match, the browser is not sent there [7]. That keeps a stray or spoofed link from sending the agent somewhere else.

• ›If runnerID can reach the browser (a hosted browser or one on your own server): Agents that drive a browser runnerID can reach, such as a hosted browser (Browserbase, Steel) or one on your own server, get the strongest mode: runnerID types the password itself.
• ›If the browser is on your own laptop: When the browser runs on your own laptop, the helper uses its local route: the password passes through your own code once to be typed, and still never reaches the model.
• ›If a hosted agent's browser is out of runnerID's reach: Hosted agents that connect over MCP get authenticator and emailed codes returned to them, and approvals and the log work the same. Typing a password into the page needs the runnerID helper next to the browser (Stagehand, Playwright or Browser Use), on a laptop or in a cloud container.

Use it on the agent's own seat on accounts you or your business control, so your agent can sign in without stopping for you. runnerID is the verified agent of a real business. It is not burner identities, not captcha solving, and not account creation for strangers. A phone number for SMS codes is coming.

• ›1. AgentMail: https://www.agentmail.to/
• ›2. Inkbox: https://inkbox.ai/ and https://inkbox.ai/docs/api/identities/manage
• ›3. Inkbox email docs: https://inkbox.ai/docs/capabilities/email
• ›4. Agent Identity: https://agent-identity.dev/
• ›5. OpenMail: https://openmail.sh/email-verification-api
• ›6. Machine Inbox: https://machineinbox.com/verification-code-api
• ›7. runnerID API docs: https://api.runnerid.torchrunner.ai/llms.txt

Browserbase and Steel belong to their owners. AgentMail, Inkbox, Agent Identity, OpenMail and Machine Inbox belong to their owners. Playwright, Stagehand and Browser Use belong to their owners. runnerID and Torchrunner are independent and are not affiliated with, endorsed by or sponsored by any of them.