Skip to main content
A mode defines what the agent focuses on: its instructions, model, capabilities, and where it can be invoked (dashboard, API, Slack).

Built-in modes

The Agent API can trigger built-in modes that have the API trigger enabled - check available_invocations in the modes list. Custom modes need the api surface. findings-investigator runs in the background with auto-approval and does not appear in the mode picker.

Custom modes

Create your own modes under Kapso Agent → Modes. Project owners and admins can create and edit them. A mode requires a name, instructions, and a model. Configurable per mode:
  • Instructions - the mode’s system prompt.
  • Model and reasoning effort - fixed for every run in this mode.
  • Surfaces - where the mode can be invoked: ui, api, slack. At least one is required.
  • Capabilities - what the agent is allowed to do (see below).
  • Resources - MCP connections, GitHub repositories, and encrypted environment variables. See Resources.
  • Sandbox - read-only or command execution, with approval, network, secret, and setup controls. See Sandbox.
The mode’s slug is generated on creation and can’t be changed afterwards.

Capabilities

New custom modes start with read_messages and read_findings. read_project_files can’t be toggled - it’s granted when you assign a repository to the mode.