What does the runnerID MCP server do?
It gives an MCP agent the tools of one runnerID identity: a vault of logins that runnerID types into the page, authenticator codes, its own inbox for emailed codes and sign-in links, saved sessions, site rules with owner approval by email link, and an append-only log. Tool results never include a password or seed. Only get_email_code and get_authenticator_code return a code, and each release is logged. [1]
How do you get an API key?
- ›1. Start free at https://api.runnerid.torchrunner.ai/signup (1 identity, 100 logins and codes a month, no card), or pick a plan at https://runnerid.torchrunner.ai: Solo is $12 a month for 1 identity, Builder $29 for 5 and Team $99 for 25, monthly through Stripe, cancel any time.
- ›2. Open the email with your sign-up link and create the account. Your API key is shown once, so store it in your password manager.
- ›3. In the owner console at https://api.runnerid.torchrunner.ai/console, add logins and authenticator keys to the identity and set its site rules. You can rotate or revoke the key there later.
How do you add it to Claude?
Add it from the terminal [2]:
claude mcp add --transport http runnerid https://api.runnerid.torchrunner.ai/mcp --header "Authorization: Bearer rid_YOUR_KEY"
Claude Code still uses an API key from the owner console.
Claude signs in through runnerID's own page. You can disconnect it any time from the owner console, under Connected apps.
- ›1. Open Settings (Customize), then Connectors [3].
- ›2. Click "Add custom connector".
- ›3. Paste https://api.runnerid.torchrunner.ai/mcp and leave the advanced fields empty.
- ›4. Claude opens runnerID's own page. Sign in and approve.
- ›5. Done. You do not paste an API key for Claude.
An Owner or Primary Owner adds the custom connector for the organization first, in Organization settings, then Connectors [3].
Custom connectors reach the server from Anthropic's cloud, not from your computer [3]. The runnerID MCP server is public, so that works.
How do you add it to Cursor or another MCP client?
In Cursor, add this to ~/.cursor/mcp.json for all projects, or to .cursor/mcp.json in a project [4]:
{ "mcpServers": { "runnerid": { "url": "https://api.runnerid.torchrunner.ai/mcp", "headers": { "Authorization": "Bearer rid_YOUR_KEY" } } } }
Cursor can read the key from an environment variable instead, for example "Bearer ${env:RUNNERID_API_KEY}" [4]. Keep the key out of shared repositories. If the account has exactly one identity, the tools use it without an identity id [1].
Cursor, Claude Code and other clients without sign-in still use an API key from the owner console.
What are the 15 tools?
- ›get_identity: Reads one identity, including its inbox address. No secrets.
- ›save_login: Stores a site login. The secret is encrypted and not returned.
- ›fill_login: Types the saved username and password into the matching page. The password is not returned.
- ›fill_code: Types the current authenticator code into the matching page. The code is not returned.
- ›save_authenticator: Stores an authenticator (TOTP) seed for a site. The seed is encrypted and not returned.
- ›get_authenticator_code: Returns the current six-digit authenticator code, for agents whose browser runnerID cannot reach. The release is logged; the seed is never returned.
- ›wait_for_email_code: Waits for an emailed code. The code is consumed and not returned.
- ›fill_email_code: Waits for an emailed code and types it into the matching page. The code is not returned.
- ›get_email_code: Waits for an emailed code and returns it, for agents whose browser runnerID cannot reach. The release is logged.
- ›open_magic_link: Opens a sign-in link only if its host is the site or a trusted subdomain. The link is not returned.
- ›save_session: Saves cookies and local storage for one site. Cookie values are not returned.
- ›restore_session: Restores a saved session into the browser. Cookie values are not returned.
- ›request_approval: Asks the account owner to approve an action, by email link.
- ›wait_for_approval: Waits for the owner's answer.
- ›list_events: Lists recent events from the log, newest first. No secrets.
We suggest adding logins in the console rather than through save_login, so the password never passes through a chat.
What works over MCP, and what needs a browser your code drives?
The fill tools type into a browser that runnerID can reach, so the agent passes the browser's address with the site.
- ›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.
How does it compare with other MCP servers for logins?
Every entry for another tool comes from that tool's own page. Not stated means its pages did not say, as of October 24, 2026.
- ›runnerID by Torchrunner: Hosted MCP server for one agent identity
- ›AgentBrowser: Hosted browser for agents, offered as an MCP server, Claude extension and more [5]
- ›Bitwarden MCP server: Local MCP server over the Bitwarden CLI [6]
- ›Autopilot MCP: Open-source browser automation MCP with Bitwarden-backed credentials [7]
- ›runnerID by Torchrunner: A browser runnerID can reach (a hosted browser such as Browserbase or Steel, or one on your own server), or a laptop browser through the helper's local route
- ›AgentBrowser: AgentBrowser's hosted browser [5]
- ›Bitwarden MCP server: Not a browser; vault actions [6]
- ›Autopilot MCP: Its own browser [7]
- ›runnerID by Torchrunner: No; typed into the page, not returned
- ›AgentBrowser: No; injected into the page, never returned to the model [5]
- ›Bitwarden MCP server: The demo shows the AI retrieving passwords and TOTP codes; Bitwarden recommends a local self-hosted LLM [6]
- ›Autopilot MCP: Not stated
- ›runnerID by Torchrunner: Yes
- ›AgentBrowser: Yes, TOTP at fill time [5]
- ›Bitwarden MCP server: Retrieves TOTP codes [6]
- ›Autopilot MCP: Not stated
- ›runnerID by Torchrunner: Yes
- ›AgentBrowser: Not stated
- ›Bitwarden MCP server: Not stated
- ›Autopilot MCP: Not stated
- ›runnerID by Torchrunner: Email link or console
- ›AgentBrowser: Approve sensitive vault requests from web or desktop [5]
- ›Bitwarden MCP server: Not stated
- ›Autopilot MCP: Not stated
- ›runnerID by Torchrunner: Append-only log
- ›AgentBrowser: Every action and vault access logged; session recording [5]
- ›Bitwarden MCP server: Not stated
- ›Autopilot MCP: Not stated
- ›runnerID by Torchrunner: Solo $12/mo, Builder $29/mo, Team $99/mo
- ›AgentBrowser: Free tier with 5 instance-hours a month [5]
- ›Bitwarden MCP server: Not stated
- ›Autopilot MCP: Not stated
Pick AgentBrowser if you want a hosted browser that comes with a vault. Pick the Bitwarden MCP server if you want an AI to manage a Bitwarden vault locally. Pick runnerID if you want logins, codes, an inbox, sessions, approvals and a log on one identity that plugs into your own agent and browser.
Questions
Is there an MCP server that stores logins and types passwords for an agent?
Yes. runnerID by Torchrunner has a hosted MCP server at https://api.runnerid.torchrunner.ai/mcp. It stores logins and authenticator keys, types them into the page, and never returns the password in a tool result.
How do I connect runnerID to Claude or Cursor?
In Claude, add https://api.runnerid.torchrunner.ai/mcp as a custom connector, then sign in and approve on runnerID's own page; no API key is pasted. In Cursor, add the URL to mcp.json with your API key as a Bearer token in the Authorization header. That key comes from the owner console.
Can I disconnect Claude from runnerID?
Yes. Claude signs in through runnerID's own page, and the owner can disconnect it any time from the owner console, under Connected apps.
Do I need to install a package?
No. The MCP server is hosted, so there is nothing to install for MCP. The Stagehand, Playwright and Browser Use helpers are separate.
Can Claude Duet use the runnerID MCP server?
runnerID makes no claim about Claude Duet. 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.
Sources, checked October 24, 2026
- ›1. runnerID API docs: https://api.runnerid.torchrunner.ai/llms.txt
- ›2. Claude Code, MCP: https://code.claude.com/docs/en/mcp
- ›3. Claude Help Center, custom connectors using remote MCP: https://support.claude.com/en/articles/11175166-getting-started-with-custom-connectors-using-remote-mcp
- ›4. Cursor docs, MCP: https://cursor.com/docs/context/mcp
- ›5. AgentBrowser: https://getagentbrowser.com/
- ›6. Bitwarden, MCP server: https://bitwarden.com/blog/bitwarden-mcp-server/
- ›7. Autopilot MCP on GitHub: https://github.com/TylerFlar/autopilot-mcp
Browserbase and Steel belong to their owners. Claude, Claude Code and Claude Duet are trademarks of Anthropic. Cursor is a trademark of Anysphere. Bitwarden is a trademark of Bitwarden, Inc. AgentBrowser, Autopilot MCP, 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.