AskCodi

Parallel feature development

Codi runs multiple agents simultaneously across different project areas, such as backend query layers and frontend chart states.

Codebase discovery and refactoring

Before work begins, Codi scans your project to build a feature map of existing code, gaps, and ready-to-run tasks.

Automated code review and approval

Changes are coordinated and brought to a single inbox for human approval, with decisions routed back to the exact engineer and task.

Multi-model orchestration

You can assign different AI models (deep reasoning, efficient models) to different tasks based on cost and complexity.

Isolated worktree management

Codi creates isolated Git worktrees for each task, keeping parallel work separate and conflict-free.

Continuous integration and testing

The platform implements and tests code changes in parallel, delivering review-ready merges.

Agent orchestration

Codi manages an entire engineering team of AI agents, assigning tasks to specific engineers, models, and isolated worktrees automatically.

Single inbox for decisions

All important approvals and decisions appear in one calm inbox; you reply once and Codi routes your decision to the correct agent and task.

Project intelligence

Codi scans your codebase before work starts, creating a living map of features, gaps, and tasks that are ready to run.

Parallel sessions

Run multiple projects from one app, with Codi coordinating parallel agents and isolated worktrees without terminal juggling.

Model economics

Codi selects the right model for each task—using deep reasoning only where it earns its cost—and tracks model spend per task.

Control room dashboard

A single view shows all projects, sessions, tasks, engineers, approvals, and model costs, including cost savings percentages.

Multi-model support

Orchestrates Claude Code, OpenAI Codex, Gemini, and 50+ other models.

Git worktree integration

Each task gets its own isolated worktree, preventing conflicts and enabling parallel development.