Two Cursor accounts.
One AgentsRoom. Zero logout.
Run a personal cursor-agent login on one project and a work login on another, in the same window, at the same time. Pin one Cursor account per project, per folder or per agent. Sign in straight from the Settings panel: no CURSOR_CONFIG_DIR to export by hand, no extra tool to install.
CURSOR_CONFIG_DIR~/.cursor-work~/.cursorOne directory per Cursor account, pointed at by CURSOR_CONFIG_DIR. Both accounts stay signed in, both run at once.
Why use several Cursor accounts
Real reasons developers run more than one Cursor account, in plain words.
Keep work and personal cleanly separated
Your employer's Cursor account stays on the work repositories. Your personal account stays on your weekend projects. No work credentials on a side project, no personal sessions recorded on the company account, and nothing to remember before you type a prompt.
Bill consumption per client
When a client gives you access to their Cursor account, the sessions for that client run on it. Their quota is spent, not yours. Each account keeps its own history, so what belongs to whom stays readable months later.
Avoid quota collisions
If your work account is on a tight plan, a long agent run on a personal project should not eat into it. Two accounts means two sets of limits and two usage histories, instead of one shared budget you keep bumping into.
Run a test account next to a prod account
Keep a second Cursor account for noisy experiments, throwaway agents and prompt drafts, and leave the main account clean. Switching between them is a pick in a list, per project or per agent.
Different orgs, different policies
A company account can belong to a Cursor team with its own privacy mode, logging and retention rules. Keeping a personal account beside it means your personal coding stays on your own terms, with credentials and configuration isolated at the directory level on disk.
Two developers, one workstation
When two developers share a machine, each signs in with their own Cursor account. AgentsRoom remembers both and starts the right one depending on the project or the agent, instead of whoever logged in last winning.
Built directly into AgentsRoom
No external switcher, no shell wrapper, no logout and login dance.
One Cursor account per project
Pin a Cursor account on a project and every agent in it inherits that account. Open a work project and a personal project in the same window: each one runs on its own Cursor login underneath.
Override per agent
Need one agent on a different account? Edit the agent, pick another Cursor account, save. The override applies to that agent only, the rest of the project keeps the project account.
One account per folder of projects
Pin a Cursor account on a sidebar folder and every project filed under it starts on that account, sub-folders included. A work folder on the work account, a personal folder on the personal one. A project, or a single agent, can still override it.
Global default account
Mark one Cursor account as the default. Every new project starts on it unless something more specific says otherwise. Change the default later and every project that pins nothing follows automatically.
Sign in inside the app
Add an account and AgentsRoom opens a small embedded terminal running cursor-agent login with the right CURSOR_CONFIG_DIR already set. Finish in the browser, then confirm: the account is ready, and nothing was typed in a shell.
Strict on-disk isolation
Each account lives in its own directory: credentials and cli-config.json. Two accounts never share an identity, and removing an account from AgentsRoom never touches another account's data.
A colour per account
Give each Cursor account its own colour. It shows as a dot on the agent avatar, on the terminal header and on the project tile, so which account is running is readable at a glance instead of being looked up.
Keep going past a usage limit
When the account an agent runs on exhausts its quota, AgentsRoom can move it onto another Cursor account you are already signed in to. One reservation: the cursor-agent CLI publishes no chat id, so the agent is relaunched cleanly on the other account instead of resuming the same conversation. Any account can be kept out of that pool with one switch.
Point at a directory you already have
An account is just a path. Let AgentsRoom create and manage one, or type the path of a Cursor configuration directory you already set up by hand, including your default ~/.cursor. Nothing is moved or rewritten.
How it works under the hood
Built on the environment variable the Cursor CLI documents itself.
CURSOR_CONFIG_DIR drives everything
The cursor-agent CLI reads its local state from one directory: credentials and cli-config.json. Default is ~/.cursor, and CURSOR_CONFIG_DIR moves it somewhere else. One directory is one signed-in account, so binding an agent to an account is binding it to a directory.
Each account is a directory
When you add an account, AgentsRoom creates a directory under its own home and signs in there. You can also point an account at any Cursor configuration directory you already have by entering its path. Both are the same thing to the CLI: a value for CURSOR_CONFIG_DIR.
Sign-in is embedded in Settings
AgentsRoom starts a small terminal running cursor-agent login with CURSOR_CONFIG_DIR set to the new directory. The browser step is the usual one. Cursor does not publish where it writes its credentials, so AgentsRoom does not guess a path: you confirm when the browser is done, rather than being told an account is ready when it might not be.
The account is resolved at launch
When an agent starts, AgentsRoom walks the cascade in order: agent override, project pin, nearest folder that pins an account, global default, then the CLI's own directory. The matching CURSOR_CONFIG_DIR is placed in the agent's environment before cursor-agent is launched.
Alternatives to AgentsRoom
Other ways to handle several Cursor accounts, and what AgentsRoom does differently.
Exporting CURSOR_CONFIG_DIR by hand
The direct route: export CURSOR_CONFIG_DIR=/some/path before launching cursor-agent. It works, and it stops working the moment you open a second terminal tab and forget which account that one is on. AgentsRoom sets the same variable for you, per project and per agent, and shows the active account in the interface.
Setting CURSOR_API_KEY
The trap almost everyone falls into. A key does not switch account: it is a fallback used when no session is active, and a session token already stored on the machine takes precedence. Anyone who has run cursor-agent login once keeps running on the first account while believing the key moved them, with no error to warn them. The directory is the only lever that actually switches accounts, and it is the one AgentsRoom drives.
Logging out and back in each time
Run cursor-agent login on the other account, then again to come back. The command replaces the cached session for the whole machine, so it is global and destructive: every other project switches with you, and any running session is disturbed. AgentsRoom keeps every account signed in at once and switches with a pick in a list.
Two machines or two OS user sessions
Some teams keep a separate machine, or a separate operating system user, per Cursor account. It is heavy and slow, and it still cannot run a work agent and a personal agent in parallel. AgentsRoom runs both at the same time, on the same machine, in the same window.
What AgentsRoom does that the others do not
AgentsRoom treats several Cursor accounts as a first-class concept: a pin per project, a pin per folder, an override per agent, an embedded sign-in that runs cursor-agent login for you, and a colour per account. Cursor accounts sit next to Claude, Codex and Grok accounts in the same panel, so a work Cursor agent, a personal Cursor agent and a Claude agent can all run side by side.
FAQ
How many Cursor accounts can I add to AgentsRoom?
There is no hard limit. An account is a directory on disk, so add a work account, a personal account, one per client and a test account if you want. They are all listed in the Settings panel and any of them can be picked for a project, a folder or a single agent.
Do I need an extra tool to run two Cursor accounts?
No. AgentsRoom drives CURSOR_CONFIG_DIR, the variable the Cursor CLI documents itself. There is no external switcher, no shell wrapper and no extra dependency. The sign-in flow runs the official cursor-agent login command in a terminal embedded in Settings.
Can two agents run on two different Cursor accounts at the same time?
Yes, and that is the point. Each agent gets its own environment, so the account is decided agent by agent. A project pinned on the work account can hold one agent overridden onto the personal account, and both run in parallel in the same window.
Does setting CURSOR_API_KEY switch my Cursor account?
No, and this is the part that surprises people. An API key is a fallback for when no session is active. Once you have signed in with cursor-agent login, the stored session takes precedence and the key is ignored, silently, with no error to tell you the switch did not happen. Separate configuration directories are the only reliable way.
What happens if I delete an account some agents still point at?
Nothing breaks. Those agents quietly fall back to the next account in the cascade, and the directory itself is left on disk in case you want to point another account at it later. You can rebind the orphaned agents whenever you like.
Are my Cursor credentials sent to a server?
No. Every account directory stays on your machine. AgentsRoom never reads, copies or transmits credentials, and there is no AgentsRoom account in the middle. The sign-in runs the official cursor-agent login CLI locally and the browser step happens between you and Anysphere, as usual.
Can I pin a Cursor account to a whole group of projects?
Yes. Sidebar folders carry their own defaults, and the Cursor account is one of them. Pin the work account on the folder holding your work repositories and every project inside starts on it, including projects added later and projects in sub-folders. The order is: agent override, project pin, nearest folder, global default, then the CLI's own directory.
Can I mix Cursor accounts with Claude, Codex and Grok accounts?
Yes. Accounts are held per provider, in the same Multiple Accounts panel. A single project can pin one Cursor account and one Claude account at once, and each agent uses the account of the provider it runs on. Nothing is shared between providers.
What exactly lives in a Cursor account directory?
The same things ~/.cursor holds for the CLI: the credentials written by the browser sign-in and cli-config.json with that account's settings. Two accounts are two of these directories, which is why they can be signed in at the same time without interfering.
What happens when one Cursor account runs out of quota?
AgentsRoom can move the agent onto another Cursor account you are already signed in to, instead of leaving it stuck until the window resets. One reservation, and it matters: unlike Claude, Codex and Grok, the cursor-agent CLI publishes no chat id, so there is nothing to resume. The agent is relaunched cleanly on the other account rather than continuing the same conversation. Accounts you would rather keep out of that pool can be excluded one by one, and the full behaviour is described on the account auto-switch page.
You may also like: Grok Multi-Account
Several Cursor accounts give you two Anysphere logins in one app. Grok Multi-Account does the same for the grok CLI, with the same pins and the same isolation. Both live in the Multiple Accounts panel, next to Claude and Codex.
See the Grok Multi-Account featureYour agent hits a usage limit. It keeps working anyway.
When the Claude or Codex account an agent runs on exhausts its quota, AgentsRoom hands the same conversation to another account you are already signed in to, and the work continues.
See how it worksFurther reading
Claude Code Keeps One Login at a Time. Here Is How to Run Several.
A field guide to running a work account and a personal account on the same machine: the one environment variable that decides which login is active, why the shell approach breaks once you have more than two terminals, and how to pin an account per project.
Antigravity CLI Keeps One Google Login Per Machine. Here Is What Works Instead.
Why you cannot alternate two Google AI Pro subscriptions on the Antigravity CLI, where it actually stores your login, what the account switchers really do to your system keyring, why a family plan does not double your quota, and the one approach that genuinely runs several accounts in parallel.
AgentsRoom Now Supports Grok Build
Grok Build, xAI's terminal coding agent, is now a first-class provider in AgentsRoom. Run it next to Claude, Codex, Antigravity CLI, OpenCode and Aider, and switch mid-conversation.
Stop logging out of Cursor. Pick the account from a list.
Run a work Cursor account, a personal one and a client account in AgentsRoom, side by side, today.
Companion app: monitor your agents on the go
Bring your own: Claude, Codex, Antigravity CLI, or other AI provider.
Push bugs and requests straight to your public backlog.
A glimpse of AgentsRoom in action.