Kapso Agent

Kapso Agent operates your project for you: it inspects APIs and logs, builds workflows, configures the inbox, manages templates and WhatsApp Flows, and investigates findings.

Kapso Agent is the project-level agent in the dashboard. It is not the agent step inside a workflow, which handles end-user conversations.

​

Where it runs

• Project home - the home page opens on the Agent view with your recent sessions.
• Agent panel - a side panel available on every project page. Pages like Logs and Findings can pre-fill a prompt with the thing you’re looking at.
• Inbox - the inbox has its own Agent sidebar.
• Slack - mention the agent in a connected channel or message it directly. See Slack integration.
• API - trigger runs programmatically and receive lifecycle events through project webhooks. See Agent API.

​

Sessions and runs

A session is a conversation with the agent. Each request inside a session is a run. Run statuses: queued, running, paused, waiting_for_approval, completed, failed, cancelled. One run executes at a time per session. Messages sent while a run is active wait in a queue - you can send a waiting message to the active run right away or remove it. Runs can be paused, resumed, or cancelled while active. For multi-step work the agent creates a plan and keeps it updated as it progresses. Simple questions and single-tool operations run without one.

​

Models

Every session starts on the mode’s default model. Change the model and reasoning effort with the selector in the chat - reasoning effort is available for models that support it. Custom modes fix the model, so the selector is locked to the mode’s configuration. In Slack, the default mode, model, and reasoning effort come from Integrations → Slack. Runs triggered through the API use the mode’s configured model - there’s no model parameter.

​

Approvals

Tool calls with side effects ask for approval before executing. Which tools need it depends on the mode’s capabilities. In Slack, the Tool approvals setting can turn approvals off for a project.

​

Settings

Configure the agent under Kapso Agent in your project:

• Resources - MCP connections, GitHub repositories, and encrypted environment variables the agent can use. See Resources.
• Modes - built-in and custom modes. See Modes.
• Sessions, Runs, API requests - execution activity, live run traces, per-run AI cost, and a 30-day history of Platform API requests made by the agent.

​

Billing

Agent runs consume your project’s AI credits. Each run and session shows its AI cost. When credits run out, runs stop with “Not enough AI credits” until you add more.

​

Next steps