Introduction

qbash lets you build Python automations that use AI models, connected services, and shared knowledge. You develop them in the task editor, and qbash runs them for you. A task can produce a report every Monday, respond to a request from another application, or become a tool someone uses from their AI assistant.

Tasks

A task contains your Python code and the prompts, documents, tests, and settings it needs. Your code decides what happens and in what order. It can call a prompt for a summary or classification, or ask an agent to investigate a question with tools.

The task editor brings those parts together. You can write code directly or ask the development agent to build and edit it with you. Test runs let you try changes before saving them.

Tasks can also be distributed to other people and applications. They can be called directly over an API, run on a schedule, or start from triggers in other systems - for example, when a new Gmail email arrives or someone posts a Slack message.

Connected services and knowledge

Connectors give tasks access to services such as a CRM, database, or messaging tool. You connect an account, then select it in the tasks that need it.

Memories hold documents that people, tasks, and agents can read and update. A folder might contain product documentation, research notes, or previous reports. Searchable folders let an agent find relevant passages across the collection.

Sharing tools with applications

An MCP server makes selected tasks, connector tools, and memory folders available to applications such as ChatGPT, Claude, Codex, and Cursor. Your team can use the same automations and knowledge from the applications they already work in.

Applications you build can connect through MCP too, using the same tools and knowledge.

Organization Settings controls who has access and which models and services are available to your team.