# Agent sign-in, accounts, and auto-stop

> Sign an agent in to its coding vendor, keep several vendor accounts and switch between them, and restart an agent that stopped after a quiet period.

Section: Your agents · Updated: 2026-11-19 · Canonical: https://nested.me/docs/agent-accounts

Some coding agents, such as Claude Code and Kimi CLI, sign in to their vendor after the agent starts. Nested shows the sign-in in chat, keeps the login on the agent, and lets the owner hold several vendor accounts. Separately, an idle agent can stop on its own to free cluster resources.

## Sign the agent in

When an agent needs a vendor login, chat shows **Sign in to use this agent** (“This agent can’t start work until it’s signed in.”).

1. Select **Sign in**.
2. Select **Open sign-in page**, or **Copy link** to open it elsewhere, and approve access.
3. If the page shows a code, paste it into **Code from the sign-in page** and select **Continue**. Some vendors update on their own once you approve.
4. If the vendor needs a browser on the agent’s machine, Nested shows **Open Desktop** so you can finish there.

> [!NOTE]
> Anyone with access to the agent uses the same vendor login. If the agent needs an API key instead, add it to the agent’s secrets and restart the agent.

## Manage several accounts

The owner manages logins in agent settings → **Runtime** → **Accounts**. One account is **Active** at a time.

1. Select **Add account**, give it a name such as “Work”, select **Sign in**, and finish the vendor sign-in. The current account stays active until you switch.
2. Select **Switch** on another account. If turns are running, choose **Wait for it to finish** (nothing is interrupted; new messages wait until the switch is done) or **Switch now** (running turns stop; their tasks stay open, and the next message continues on the new account).
3. Use **Rename**, **Sign in again**, or **Remove**. When you remove the active account, choose the account to switch to, or **Sign the agent out**.

If the new account cannot sign in, Nested keeps the previous account active and tells you.

## When an agent stops after a quiet period

To free resources, an agent stops after a quiet period with no work — ten minutes by default. Chat then shows **This agent is stopped** — “It stopped after a quiet period to free resources. Start it to continue.”

1. Select **Start agent**.
2. Wait for **Starting the agent…**. This can take a few minutes. You can send messages once it is ready.

An agent that was stopped by hand shows “It was stopped by hand” instead; **Start agent** works the same way.

## Example

Your team shares a Claude Code agent. Its owner adds a second account named “Overflow”. When the first account hits a limit, the owner selects **Switch**, chooses **Wait for it to finish**, and new tasks continue on “Overflow”.

## Questions

### Does everyone need to sign in to a shared agent?

No. The vendor login lives on the agent, so everyone who uses it shares the same login.

### Why did my agent stop by itself?

It stopped after a quiet period to free cluster resources. Select **Start agent** to bring it back; your tasks and workspaces are kept.

## Related

- [Deploy an agent](https://nested.me/docs/deploy)
- [Agent settings](https://nested.me/docs/agent-settings)
- [Agent runtime](https://nested.me/docs/agent-runtime)
