Settings & Preferences

NimbleBrain organizes settings by scope — what each setting affects:

• You — your personal profile (theme, display name, timezone). Lives at /profile, reached from the avatar menu at the top-left of the sidebar.
• This Workspace — settings for the workspace you’re currently focused on (its name, members, instructions, connectors and their settings, skills). Lives under /w/<id>/settings/*.
• Organization — admin-only settings that affect every workspace (model defaults, the workspace roster, deleted-workspace archives, user management, org-wide usage). Lives under /org/*.

Open profile or sign out from the avatar menu at the top-left. Open workspace and org settings from the workspace navigator and the workspace settings nav.

Your profile (/profile)

Reach your profile any time from the avatar menu at the top-left of the sidebar — click your avatar, then Profile settings. The profile surface is tabbed.

General (/profile/general)

• Display name — how your name appears to teammates.
• Email — your login email (read-only).
• Role — your org-level role (read-only).
• Timezone — used for scheduling and time-related responses.
• Your model — the model your new conversations start on, or follow the organization’s default.
• Theme — System (follows your OS), Light, or Dark.

Each field saves as you change it: your name when you leave the field or press Enter (Escape undoes the edit), everything else when you pick. A save shows a notice with Undo; one the server refuses stays on the field with the reason, Retry and Revert.

Skills (/profile/skills)

Your personal skills — prompt-composition guidance the agent picks up across all your workspaces.

The avatar menu also has a Sign out action.

This Workspace

This section reflects the workspace you’re currently focused on. Switch workspaces from the navigator at the top-left of the sidebar; the workspace settings update to match.

General (/w/<id>/settings/general)

The workspace ID sits in the page’s top-right corner; click it to copy.

• Name — what the workspace is called in the sidebar, the switcher, and everywhere else. An admin member of this workspace or an organization admin can rename it; it saves when you press Enter or click away, with Undo in the notice. Renaming never changes the workspace’s ID or URL.

Members (/w/<id>/settings/members)

Who belongs to this workspace and their role. Every member can see the list.

An admin member of this workspace, or an organization admin, can also manage it:

• Add member — enter the email of someone who already has an account in your organization, and pick their role.
• Role — change a member’s role from the list; it saves at once, with Undo in the notice.
• Remove — asks before taking someone’s access away. Removing yourself leaves the workspace.

The workspace’s last admin can’t be demoted or removed, so a workspace always has someone who can manage it.

Connectors (/w/<id>/settings/connectors)

Browse and configure the workspace’s connectors — point-and-click OAuth integrations and managed connectors. Install a connector here to make its tools available to the agent in this workspace.

Open a connector to see its settings page: its status, connection, secrets, and tool permissions, then any settings the connector provides itself, under its name. Every member can open the page; only an admin member of this workspace can change the host’s sections.

Some connectors support custom instructions of their own — saved guidance the agent picks up whenever you’re using that app. When supported, the editor appears in the connector’s own section of its settings page; the connector owns storage and validation, and the platform includes the saved text in the agent’s system prompt on every turn. See the app-author docs for the other side of this.

Skills (/w/<id>/settings/skills)

Everything shaping your agent in this workspace, in one list, grouped into tiers by where it comes from: Yours (the skills you own here), You (your personal skills, which ride into every workspace), Organization (org-wide policy), and System (built-in). A segment filter above the list slices it to any one tier. Each row states at rest how the skill loads:

• Always on · every conversation — composed into the agent’s instructions on every turn.
• On tool match <glob> — loaded when a matching tool comes into play.
• On trigger “<phrase>” — loaded when the phrase appears in the conversation.
• When the agent asks for it — listed in the skill catalog by its description, and loaded when the agent decides the job needs it.

Toggle any skill you own on or off, or expand a row to read and edit it. Tiers you don’t edit here are read-only (shown with a lock) and link to where they’re managed — personal skills to your profile, organization skills to org settings, system skills are fixed. + Add a skill writes to the tier you own on this page, when you click Create. Editing a skill that exists has no Save button: each field saves on its own, as described in Editing a skill in settings.

MCP (/w/<id>/settings/mcp)

The workspace’s MCP URL, <origin>/mcp/<workspaceId>, with a copy button. Add it as a custom connector in an MCP client (Claude, Claude Code, Cursor) to reach this workspace’s tools from there. Every member sees it; the client signs in as you and reaches the workspace only while you are a member. See Connecting External Clients.

Organization (admin only)

Visible only to organization admins. Hidden from regular members. These settings live under /org/*, outside any workspace URL.

Workspaces (/org/workspaces)

The org-wide workspace roster: create and delete workspaces, and see each one’s connectors and metadata. Open a workspace here to add or remove its members without joining it, including seating an admin in a workspace that has none. Change an existing member’s role from that workspace’s Settings → Members.

Deleting a workspace asks first. It disappears for everyone in it and its connectors are disconnected, which can’t be undone; its conversations and files are kept under Archives until an admin purges them.

To edit a workspace’s instructions, switch into it via the navigator and use This Workspace → General — and note that editing requires admin membership of that workspace.

Users (/org/users)

Manage user accounts at the organization level: create a user with a Member or Admin role, edit a user’s display name, email, and role, deactivate one, or restore a deactivated one. Edits save as each field changes. See User Management.

Models (/org/model)

Choose which AI models the agent uses across every workspace. Each field saves as you change it: a dropdown when you pick, a number when you leave the field or press Enter (Escape undoes the edit). A save that applies takes effect at once and shows a notice with Undo; one the server refuses stays on the field with the reason, Retry and Revert.

• Default model — primary model for chat.
• Fast model — cheaper, faster model for quick auxiliary tasks. Still give it a large context window, not just low latency/price — a fast request carries the same tool surface as a normal one, so a small-context model starves conversation history (details).
• Reasoning model — most capable model for complex analysis.

Plus runtime limits — max iterations and max input/output tokens — and the Extended thinking control.

The extended thinking modes are:

• Default — adaptive for reasoning models, off otherwise.
• Off — never reason.
• Adaptive — the model decides per call.
• Enabled — always reason. Enabling this reveals a Thinking Budget Tokens field (minimum 1024), which counts toward Max Output Tokens.

Reasoning is billed as output tokens, and the agent’s reasoning shows up in chat as a collapsible Thought chip. See Chat.

Caution

Model configuration is platform-wide. Changes affect every workspace and every user immediately. Restrict to org admins.

Usage (/org/usage)

Organization-wide usage analytics. All dates are UTC.

• Filters — narrow everything on the page by Workspace, User, and Model, over a Range: month to date, the last 7 or 30 days (each counting today), last month, or a custom from/to.
• Summary cards — total cost, chat cost, task cost, and tokens. Click a card (or focus it and press Enter) to see its detail in place: cost by input, output, and cache; conversations or task runs; LLM calls; cache hit rate. Chat and task spend are shown apart, since a task is not someone chatting. Cost excludes any call whose model has no known price, and says so when it does.
• Group by — the daily cost chart stacks, and the table breaks down, by Model, Workspace, User, or Origin (chat versus task).

Note

Usage is reported at the organization level, on this page. Filter or group by workspace to see one workspace’s spend; there is no per-workspace usage tab.

Archives (/org/archives)

Deleting a workspace archives its data rather than destroying it. This tab lists every archive: the workspace’s name and ID, the size on disk, and when it was archived. An archive whose workspace record is missing or unreadable is still listed, with its name and ID shown as unknown.

Purge permanently removes one archive from disk. It asks first, stating the size, and cannot be undone. Nothing is purged automatically — archives stay until an admin purges them, so check this tab when disk space runs low.

Custom-instruction layers

Custom instructions stack from broad to narrow, and the agent reads them on every turn:

1. Organization policy applies everywhere.
2. Workspace skills set to always on (This Workspace → Skills) layer on top for the focused workspace. Use one for team conventions: tone for shared documents, default triage priorities, citation policies.
3. App custom instructions (This Workspace → Connectors → the connector) apply when you’re using that app.

Sidebar layout

The sidebar collapses with the toggle on its right edge (or ⌘B / Ctrl-B). Collapsing hides labels but keeps icons; expanding restores the full nav. Your avatar and the workspace navigator sit at the top-left; clicking the avatar opens a popover with Profile settings and Sign out.