Agents & Modes
AdaL starts as a coding agent by default: ask it to inspect your project, make changes, run tests, and help you ship.
When the task calls for a different capability, use /agent to switch modes. AdaL Engineer is a separate workflow for more autonomous work, with loop engineering built in.
Quick Start
adal
Use the default coding agent for everyday implementation, debugging, and review work.
/agent
Open /agent at any time to select a specialized agent. Use ↑/↓ to navigate and Enter to select. Press Esc or Ctrl+C to close it.
Choose the Right Agent
| Agent | How to start | Best for |
|---|---|---|
| AdaL | adal | Default coding agent for editing files, running commands, debugging, and test/fix loops |
| Deep Research | /agent → Deep Research | Multi-source investigation, technical decisions, and structured research |
| Browser Use | /agent → Browser Use | Real browser interaction, UI verification, and web workflow automation |
| Self-Review | /agent → Self-Review | Risk analysis and pre-merge review |
AdaL: The Default Coding Agent
Start a normal AdaL session when you want to work with the agent directly:
adal
Describe the outcome you want. AdaL can explore the codebase, edit files, run tests, use tools, and explain the result.
Use /agent when a task benefits from a specialized mode:
- Deep Research for evidence-heavy technical research.
- Browser Use for validating a live site in a real browser.
- Self-Review before merge to surface risk and missing validation.
AdaL Engineer Is a Separate Workflow
AdaL Engineer is not selected through /agent. Start it explicitly when work benefits from planning and parallel delegation.
adal --agent-mode engineer
Give it a goal, then let its supervisor plan and coordinate workers across implementation, testing, and review.
Use AdaL Engineer for:
- Multi-file refactors across an unfamiliar codebase
- Bugs that span services or components
- Features that need planning, implementation, tests, and review
- Several independent tasks that can progress in parallel
AdaL Engineer is currently available in the CLI only and is not supported in the Desktop app.
Headless Use
For scripts or non-interactive workflows, select a mode with --agent-mode:
adal -q "analyze this architecture" --agent-mode research
adal -q "review my staged changes for risks" --agent-mode self-review