Product guide
Do the work. Keep the controls.
Everything your team needs to set up a project, run an agent, approve a change and operate the workspace. The API reference is generated from the OpenAPI contract.
Getting started 4
OverviewAI Engineering Platform is a workspace where AI agents execute the software lifecycle and people keep the decisions that matter.QuickstartThis guide runs the platform on one machine with Docker Compose. It is the fastest path to a first approved, merged task.Key conceptsThe platform implements an AI Software Development Lifecycle ASDLC: a sequence of phases, artifacts and gates. The model is replaceable; the process is the product.Release notesThe changes that reach you, newest first. The full history is in CHANGELOG.md.
Build & ship 5
Projects and blueprintsA blueprint is the only entry point for creating a project. It captures the decisions that should not be reinvented for every repository.Tasks and boardThe board is the source of truth for what the team is doing. It is deliberately not a prompt box.Agent runs and approvalsA run is one agent execution. It has a type, a frozen context, a budget and an event stream. The run view is where a team sees what the agent did and decides what happens next.Pull requests and mergesThe platform opens a pull request when a development run is ready. The PR is the hand-off between the agent's diff and the team's repository rules.Git providersThe platform talks to your repository host with a token. That token pushes
Quality & insights 2
Administration 5
Users and accessWorkspace access is role-based and auditable. The same role model applies to the UI and the API.Security and auditThe model is not the security boundary. The runtime, the approval gates and the audit trail are.Agents, models and BYOKThe platform is model-agnostic. A project chooses which agents and providers it may use; the platform records the choice on every run.Backups and recoveryBackups are a product surface for operators, not a cron job hidden on a server.TroubleshootingStart with the run view and the event stream. Most failures have a specific event before the generic error.