Tools

The Tools page is where you manage which API actions agents in your workspace are allowed to request.

Every tool call an agent makes must reference a tool that is registered, approved, and active. If a tool does not meet all three conditions, the call is blocked - regardless of what policies the agent carries.

The three tabs

Active

Approved tools that agents can use. You can enable or disable any active tool using the toggle on its row.

Disabling a tool stops all agents from calling it immediately, without changing any policies. This is useful when a third-party API has an outage, a security issue is being investigated, or a tool needs to be temporarily suspended. The tool stays in your registry and can be re-enabled at any time.

Pending

Tools that have been registered but not yet approved. Agents cannot call a pending tool. Review each pending tool and either approve it (moving it to Active) or reject it with a reason.

New tools appear here when an agent attempts to call a tool that is not yet in your registry - KeyRunner captures the request and queues it for review rather than silently blocking it.

Rejected

Tools that have been explicitly rejected. These are not available to any agent. The rejection reason is stored so future reviewers understand the decision.

What each tool shows

Each tool entry displays:

• Tool name - the exact identifier agents use when making a call (for example, github_create_issue)
• API collection - which product or API this tool belongs to (for example, GitHub, Stripe, Salesforce)
• Category - the type of action: data retrieval, communication, search, or other
• Requested by - which agent first requested this tool
• Redaction rules - which data fields are automatically hidden from logs

Redaction rules

Some API responses contain sensitive data - API keys, card numbers, passwords - that should never appear in logs. Redaction rules let you specify which fields KeyRunner should automatically hide.

When a tool has redaction rules set, KeyRunner scrubs the listed field names from the API response before writing the log entry and before returning the result to the agent. The original value is never stored anywhere.

Common fields to redact: api_key, card_number, password, token, secret, ssn.

Redaction is configured per tool when a tool is registered or edited. It cannot be bypassed by the agent.