runnerID by Torchrunner

Logs your AI agent into websites, catches verification codes, and asks you before anything risky.

Hosted MCP Server

npx add-mcp 'https://api.runnerid.torchrunner.ai/mcp'

Installs into Claude Code, Codex, Cursor and more

Documentation

Cross-platform agent identity

One identity that follows the operator across tools. runnerID logs your agent in, catches the code, keeps the session, and asks you before anything risky.

Passwords shown to the model

Never

Works with

MCP

Every action

Logged

siteSupplier portal · allowed for this agent

passwordTyped by runnerID. The model never sees it

codeSix-digit code caught in the agent's inbox, typed

sessionSaved, so the agent stays signed in next run

waitNew payee added. Waiting for your yes

logEvery step recorded. One switch revokes it all

Every wall between your agent and the work

Passwords, without exposure

runnerID types the login into the page. The model never sees it, so it cannot repeat it in prompts, tool output or logs.

Six-digit codes

Authenticator codes are made in the vault and typed for your agent. No phone in the loop.

Check your email

Every agent gets an inbox that catches the code or magic link and uses it.

New device prompts

Saved sessions keep your agent signed in between runs.

Ask before acting

You set which sites it may touch and what needs your approval. Approve or deny from an email link.

A record of everything

Every fill, code and approval is logged. One switch revokes it all.

How it works

Step 1

Create an identity

Save its logins and authenticator codes once. They are encrypted with a key for that identity alone.

Step 2

Point your agent at runnerID

One MCP connector, or a helper for Stagehand, Playwright or Browser Use.

Step 3

runnerID fills the page, catches the code, and pauses for you when an action needs approval.

When your agent gets stuck

Your agent can hand runnerID the step it can't finish.

Sometimes your agent can't get into a site, can't get past a login, or can't fill in a form. Soon it will be able to hand that one step to runnerID.

runnerID does the step for you, on your own accounts, in its own browser. Where a site offers a staff or delegate seat, your identity uses its own seat, not your password.

Before anything that sends, pays, deletes or changes settings, runnerID asks you first. Nothing happens on that step without your yes.

Then it hands the result back to your agent so the task can carry on. Every step is logged, and if a run stops, you and your agent get a clear note of where it stopped and why.

How the handoff will work

Hand off the step

Your agent says what to do, on which site, and what to bring back.

Your approval first

Sending, paying, deleting and changing settings wait for your yes.

Where it stopped

Every run is logged, and a run that stops says where and why.

Cross-platform agent identity

What runnerID is

runnerID by Torchrunner is cross-platform agent identity: one identity that follows the operator across tools. runnerID logs your AI agent into websites, catches the verification code, keeps the session, and asks you before anything risky. Your agent never sees a password. Each identity holds a vault of logins that runnerID types into the page, authenticator codes, its own inbox for codes and sign-in links, saved sessions, site rules with owner approval by email link, and an append-only log with revoke.

›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.

One identity that follows the operator across tools

Claude Duet, Grok Bot, OpenAI Dots and Meta Muse each work inside their maker's own app and account. runnerID is the operator's own identity layer instead: the same logins, codes, inbox and saved sessions, whichever agent is signing in.

runnerID does not claim support for, or an integration with, any named product.

Plans

Pay monthly and cancel any time. Every plan has every feature. Plans differ by how many agent identities you run.

Solo

1 identity

$12/ mo

One agent identity with its own logins, codes, inbox and log.

Passwords typed into the page, never shown to the model
Authenticator and emailed codes
An inbox for each identity
Saved sessions between runs
Owner approvals by email link
A full log and one switch to revoke
Owner console, MCP connector and helpers

What is included	Solo
Solo	1 identity, $12 a month
Builder	5 identities, $29 a month
Team	25 identities, $99 a month
Logins per identity	Unlimited
Approvals and the log	On every plan
Torchrunner subscribers	Included, one identity per engagement

What happens after I pay?

You get an email with a sign-up link that lasts 7 days. It creates your account and shows your API key once. Then connect your agent with the MCP connector or a helper, and manage identities, approvals and the log from the owner console.

How do I cancel?

Any time, from the Manage billing link at the bottom of this page or the link in your receipt. Your plan runs to the end of the month you paid for. Plans are month to month, with no contract. After that month ends, the account pauses, and your data is kept for 30 days so you can export it.

Does it work with hosted agents?

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.

Running more than one agent?

Builder covers 5 identities and Team covers 25. Same features, more identities.

Questions

What is runnerID?

runnerID is a login service for AI browser agents, made by Torchrunner. It types saved passwords into the page so the model never sees them, fills authenticator and email codes, keeps sessions between runs, and asks the owner before risky actions. It connects through MCP or helpers for Stagehand, Browser Use and Playwright.

Does my agent see my passwords?

No. When runnerID can reach the browser, it types the password straight into the page, so it never enters the model, its tool output or its logs. 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.

Which agents does it work with?

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. A browser on your own laptop uses the helper local route. Works with Claude, Cursor, Stagehand, Browser Use and Playwright.

Is this for bots pretending to be people?

No. runnerID is the verified agent of a real business, working on accounts the owner controls.

Where are secrets kept?

Passwords and authenticator seeds are encrypted with a key per identity and decrypted only to be typed. The laptop route hands the password to your own code once. Email codes are kept briefly so they can be used, and logs never record secrets.

Does runnerID train AI models on my data?

No. Your data is yours. We never use your content to train AI models: not your records, documents, emails, passwords, codes, or anything on the pages your agents visit. We do learn from how tasks get done, such as which steps a sign-in needs or where a site asks for a code, with your content and personal details removed, so agents get better at the work. You can turn this off any time in your settings, and we never sell it or share it for anyone else to train on.

Can I use runnerID with Claude Duet, Grok Bot, OpenAI Dots or Meta Muse?

What does runnerID save?

Fewer tokens, because saved sessions keep the agent signed in between runs. Time, because codes are made and typed for it and emailed codes are caught in its own inbox. Friction, because you approve risky actions from an email link. And you keep control: the identity is yours, with a key per identity, and one switch revokes it all.

How do I get access to runnerID?

Pick a plan under Pricing on this page. After you pay, you get an email with a sign-up link that creates your account and shows your API key once.

How do I stop pasting passwords into AI agent prompts?

Move the password out of the prompt and into a vault that types it into the page for the agent. With runnerID, you add the login once in the owner console, and it is never shown back. When the agent reaches the login page, it asks for the login by site name, and the password is typed into the page. The model never receives it, so it cannot repeat it.

Is it safe to give an AI agent my password?

Putting the password in the model's context is the risky part, so keep it out entirely. runnerID stores the login in a vault and types the password into the page; the model never receives it. Site rules limit which sites the agent may use. Risky actions wait for your approval by email link. Every action goes into an append-only log, and revoke shuts it all off at once.

How do I give an AI agent its own email inbox?

Give the agent an identity with its own inbox address. Each runnerID identity has one, and it catches emailed verification codes and sign-in links. Use that address as the email on accounts you control, or as the address a site sends codes to. When a code arrives, the agent can have it typed into the page, and hosted agents that connect over MCP get emailed codes returned to them.

Can an AI agent use email sign-in links (magic links) safely?

Yes, if the link lands in an inbox built for the agent and is checked before it opens. Each runnerID identity has its own inbox address that catches emailed sign-in links. The agent waits for the link, and it is opened in the browser only when the link's host matches the site the agent is signing in to.

How do I require human approval before an AI agent acts?

Set rules for each identity: which sites it may use and which actions need you. When the agent reaches a risky step, it asks for approval, and runnerID emails you a link to approve or deny. You can also approve or deny from the owner console. Every request and decision goes into the append-only log. Approvals work the same for hosted agents that connect over MCP.

What happens when my agent can't finish a step?

Coming to runnerID: your agent will be able to hand the stuck step to runnerID, such as a login, a code or a form on a site it can't reach. runnerID does that step for you, on your own accounts and in its own browser, asks you before anything that sends, pays, deletes or changes settings, and hands the result back to your agent. Every step is logged, and if a run stops, you and your agent are told where and why.

Which agents will be able to hand runnerID a step?

Coming to runnerID: agents that can call an outside tool you connect them to. What each one can do depends on its maker and your plan. Claude lets you add custom connectors on every plan (one on Free). Gemini on the web lets adults in the US with a personal Google Account add custom apps. runnerID does not claim support for, or an integration with, any named product.

Will runnerID use my password when it does a step for me?

Coming to runnerID, and the rule is simple: not where a site forbids sharing it. Where a site offers a staff, delegate or service provider seat, your identity gets its own seat with only the access you give it, and you can remove it on the site any time. Where the site has its own app connection, runnerID uses that. Your own password is used only on sites whose terms allow it.

What does runnerID cost?

runnerID has three monthly plans through Stripe, and every plan has the same features. Solo is $12 a month for 1 identity. Builder is $29 a month for 5 identities. Team is $99 a month for 25 identities. An identity is one set of logins, codes, inbox, sessions and rules that your agents use. You can cancel any time.

What counts as an identity?

An identity is one set of logins, codes, inbox, sessions and rules that your agents use. In runnerID it holds a vault of logins, authenticator codes, its own inbox address, saved browser sessions, site rules with owner approval, and an append-only log. Several agents can use the same identity. Solo includes 1 identity, Builder 5 and Team 25.

What does the owner console do?

The owner console at https://api.runnerid.torchrunner.ai/console is where the account owner runs runnerID. You sign in with an emailed link. From there you create and revoke identities, add logins and authenticator keys (never shown back), set site rules, approve or deny requests, rotate or revoke API keys, and read the activity log.

What happens after I pay for runnerID?

How do I connect runnerID to Claude or Cursor?

Add it as an MCP connector. The hosted MCP server is at https://api.runnerid.torchrunner.ai/mcp, and you send your API key as a Bearer token. That works in Claude, Cursor and any MCP client. 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.

How does an AI agent stay logged in between runs?

Save the browser session after the first login and restore it at the start of the next run. runnerID saves sessions per site for each identity, so the agent starts signed in instead of facing the login page and the code again. If a site still asks, the saved login and authenticator code can be typed for it. Revoke shuts it all off at once.

Is runnerID a way to get around website rules?

No. runnerID is the verified agent of a real business, working on accounts the owner controls. It is not burner identities, not captcha solving, and not account creation for strangers. It signs in to accounts you already own, with site rules you set, your approval for risky actions, and an append-only log you can read in the owner console.

Does runnerID support passkeys or SMS codes?

Not yet. Both are coming: passkeys, and a phone number for SMS codes. Also coming are agent payment cards, a fixed IP, and signed agent traffic. Today runnerID handles saved passwords typed into the page, authenticator (TOTP) codes, emailed codes and sign-in links in the identity's own inbox, saved sessions, owner approvals and the log.

How do I handle 2FA in a Stagehand agent?

Install the helper with npm install @runnerid/stagehand. Save the site's login and authenticator key once in the runnerID owner console, for the agent's own seat where the site offers one; neither is shown back. Your Stagehand code calls rid.login(site) and rid.code(site), and the password and the current code are typed into the page. The model never receives either value. For emailed codes, use the identity's own inbox address and rid.emailCode(site).

How do I log in with 2FA in Browser Use?

Install the helper with pip install runnerid. Keep the login and the authenticator key in runnerID, for the agent's own seat where the site offers one, not in the task prompt. Your Browser Use script calls rid.login(site) and rid.code(site) before or during the run, and the password and the current code are typed into the page. The model never receives the password. Emailed codes go to the identity's own inbox, and rid.email_code(site) types them.

How do I automate 2FA logins in Playwright for an AI agent?

Install the helper with npm install @runnerid/playwright. Store the login and the authenticator (TOTP) key in runnerID, for the agent's own seat where the site offers one, then call rid.login(site) and rid.code(site) when your script reaches the login page. The password and the current code are typed into the page, so your prompts and the model never hold them. Save the browser session after the first login with rid.save(site), so later runs start signed in.

Not ready to buy yet?

Tell us where your agent gets stuck and we will reply.

Give your agent its own identity

Solo is $12 a month for one identity. Cancel any time.

About the names

Browserbase and Steel belong to their owners. Claude and Claude Duet are trademarks of Anthropic. Grok Bot is a product of xAI and Cursor. OpenAI, ChatGPT, ChatGPT Atlas and Dots are trademarks of OpenAI. Meta and Muse are trademarks of Meta. Gemini and Chrome are trademarks of Google. Perplexity and Comet are trademarks of Perplexity. Other product names belong to their owners. runnerID and Torchrunner are not affiliated with, endorsed by or sponsored by any of them.