Core Concepts
Obi is organized around a few building blocks that work together.
Workspaces
A workspace is your top-level container. Everything you build in Obi lives inside a workspace. You can invite team members, connect integrations, and manage billing at the workspace level. If you work across different teams or clients, you can create separate workspaces for each.
Apps
An app groups related agents, squads, and tools together. Think of an app as a project folder for a specific automation you want to build. For example, you might create one app for customer support automation and another for code review.
Agents
An agent is a single AI worker. You give it a prompt (instructions for how it should behave), choose a language model, and optionally attach tools and variables. Agents handle one part of a larger workflow.
Squads
A squad is a multi-agent workflow. Using a visual editor, you connect agents together into a sequence or branching flow. When a squad runs, each agent executes in order, passing results along to the next. Squads are where individual agents come together to accomplish complex tasks.
Triggers
A trigger starts a squad automatically when something happens. Obi supports three types of triggers: incoming emails, API calls from external systems, and delegated tasks from Linear. Triggers let you run your workflows without manual intervention.
Executions
An execution is a single run of a squad. You can start one manually or let a trigger create one. While an execution is running, you can watch its progress in real time, see what each agent is doing, and provide input if the workflow pauses to ask for it.
Projects
A project connects Obi to a code repository. It stores context and instructions that agents can reference when working on tasks related to that codebase. Projects are tied to a git integration like GitHub.
Integrations
Integrations connect Obi to external services. GitHub and Linear are the primary integrations, but Obi also supports tool providers like Composio for adding hundreds of additional capabilities to your agents.
Memories
Memories are persistent key-value pairs stored at the workspace level. Agents can read and write memories during executions, making it possible to carry information across separate runs.