Toggle tool categories on or off, view individual tools, configure browser access, and set tool-level restrictions for this agent.
Tools are the individual functions your agent can call — sending emails, reading files, searching the web, posting to social media, managing calendars, and more. They are organized into categories (e.g., "Email Tools," "Google Workspace," "Messaging").
This page lets you turn entire categories on or off with a toggle switch, see exactly which tools are in each category, and apply filters to find what you need. Think of it as a control panel for your agent's capabilities.
Tool data is fetched from GET /bridge/agents/:id/tools, which returns an array of category objects, each containing:
id — Category identifier used in toggle requestsname, icon, description — Display metadataenabled — Whether the category is currently ontoolCount — Number of individual tools in this categorytools — Array of tool name strings (shown on expand)alwaysOn — If true, the category cannot be disabledrequiresOAuth — OAuth provider needed (e.g., "google")requiresIntegration — Whether an external service must be configuredplatformUnavailable / platformMessage — When the platform isn't supportedisAvailable — Whether prerequisites (OAuth, integration) are satisfiedToggling sends PUT /bridge/agents/:id/tools with a body mapping category IDs to booleans. The response is optimistically applied client-side, with a full reload on bulk operations.
Tools are grouped into logical categories. Each category represents a related set of functionality. Examples include:
Each category has an icon that is mapped from server-side emoji strings to custom SVG icons for consistent rendering across platforms.
Some tool categories are marked Always On and cannot be disabled. These are core capabilities the agent needs to function — things like memory management, basic utilities, and internal communication. They appear with an "Always On" badge and no toggle switch.
Categories like Google Workspace require an OAuth connection before they can be used. If OAuth hasn't been configured:
The filter tab bar provides quick views:
google_ prefixlocal_ prefix (file, shell, etc.)enterprise_ prefixClicking a category card (not the toggle) expands it to show the individual tool names within. Tools are displayed as monospaced badges — green-tinted when the category is enabled, gray when disabled. This lets you audit exactly which functions the agent gains access to.
Two bulk action buttons sit in the stats bar:
Below the tool categories, a Browser Configuration card (from BrowserConfigCard) lets you configure the agent's headless browser settings — user agent, viewport, proxy, and other browsing parameters. This controls how the agent interacts with web pages.
The Tool Restrictions card (from ToolRestrictionsCard) provides fine-grained control beyond category-level toggling. Here you can block specific individual tools, set execution policies, or configure safety guards on particular tool calls.
Only enable the tool categories your agent actually needs. A customer support agent doesn't need local system tools. A research agent doesn't need messaging. Start minimal and add capabilities as needed.
Expand a category to see its individual tools before toggling it on. Some categories contain powerful tools (e.g., shell execution, file deletion) that deserve careful consideration.
Connect your Google account in the Email tab before enabling Google Workspace categories. Enabling them without OAuth means the tools exist but can't function — potentially confusing the agent.
Periodically use the "Enabled" filter to review what's active. Use "Integrations" to check which external services are connected. This is especially important for security-conscious deployments.
Tool toggles control availability; the Permissions tab controls risk levels and approval requirements. Use both together — enable a category here, then set approval requirements for its risky tools in Permissions.
The category's platform isn't supported in the current environment. Check the platformMessage for details — it may require a specific OS, runtime, or service that isn't available on this server.
Check the browser console for API errors from PUT /bridge/agents/:id/tools. Common causes: the agent engine is unreachable, or the agent ID is invalid. Try refreshing the page.
The tool list may be cached. Refresh the page to re-fetch from the server. If it persists, check that the OAuth token in the Email tab is valid and hasn't expired.
This is expected. Always-on categories are already enabled and can't be toggled. The bulk action only affects optional categories.
The stats bar shows the sum of toolCount across enabled categories. If a category has alwaysOn: true, its tools are always counted as enabled even though they don't appear in the toggle list.
When an agent belongs to a client organization, tool access is governed by that organization's policies: