# Komos

> Komos is an AI automation platform. Users create reusable automation tasks that can browse websites, call 100+ API integrations, process data with AI, send emails, parse documents, and more. Tasks run in managed cloud sandboxes or self-hosted desktop runners.

## Instructions for LLMs

- Prefer the MCP server for agent-to-agent interaction (run tasks, check results, talk to Moss)
- Use the REST API for programmatic task execution from application code
- Moss (ask_moss MCP tool) can create tasks, configure integrations, set up schedules, and build automations on behalf of users
- Tasks are the primary entity; they contain node graphs with browser automation, data processing, integrations, and more
- Do not confuse Komos with Zapier/Make; Komos automates at the browser level using AI vision, not just API connectors

## MCP Server

- Endpoint: https://api.komos.ai/public/v1/mcp/
- Transport: Streamable HTTP
- Auth: Bearer token (personal API key, prefix kms_)
- Tools: list_tasks, get_task, run_task, get_run (with include_logs), list_runs, list_skills, get_skill, list_credentials, list_integrations, ask_moss

## Docs

- [Tasks](https://docs.komos.ai/): Core automation entity. Reusable workflows with browser automation, integrations, data processing, and AI agents.
- [Moss](https://docs.komos.ai/moss): AI automation engineer built into every account. Creates tasks, monitors runs, fixes issues.
- [MCP Server](https://docs.komos.ai/mcp-server): Connect AI coding agents to Komos. Setup for Claude Code, Claude Desktop, Codex, Gemini CLI.
- [Agent Action](https://docs.komos.ai/agent-action): General-purpose AI agent node for browser automation, data processing, integrations, and more.
- [Credentials](https://docs.komos.ai/credentials): Manage login credentials and MFA for automated browser sessions.
- [Integrations](https://docs.komos.ai/integrations): 100+ native API integrations. AI routes to the right one automatically.
- [Skills](https://docs.komos.ai/skills): Reusable instruction sets that extend task capabilities.
- [Environments](https://docs.komos.ai/environments): Cloud sandboxes and self-hosted desktop runners (Windows/macOS).
- [Execution](https://docs.komos.ai/execution): Monitor runs, view logs, retry failures, export outputs.
- [API & Webhooks](https://docs.komos.ai/api-webhooks): REST API for triggering task runs and receiving webhook notifications.
- [SDKs](https://docs.komos.ai/sdks): Python and TypeScript SDKs for programmatic task execution.
- [Organizations](https://docs.komos.ai/organization): Team management, roles, and permissions.

## API Reference

- [Introduction](https://docs.komos.ai/api-reference/introduction): REST API overview, authentication, base URL.
- [Queue a task run](https://docs.komos.ai/api-reference/task-runs/queue): POST /tasks/{task_id}/runs — start a task execution with optional inputs.
- [List task runs](https://docs.komos.ai/api-reference/task-runs/list): GET /tasks/{task_id}/runs — list recent runs for a task.
- [Get a task run](https://docs.komos.ai/api-reference/task-runs/get): GET /task-runs/{run_id} — get status, outputs, and details.
- [Get run logs](https://docs.komos.ai/api-reference/task-runs/logs): GET /task-runs/{run_id}/logs — step-by-step execution logs.
- [Cancel a task run](https://docs.komos.ai/api-reference/task-runs/cancel): POST /task-runs/{run_id}/cancel — abort an in-flight run.
- [Webhooks](https://docs.komos.ai/api-reference/webhooks/overview): Receive run.completed, run.failed, and other events via HTTP callbacks.