Getting Started

Quickstart

This guide walks you through the core Exponential workflow: create a task, assign an agent, approve a plan, and review the output. You'll see how the plan-first model works in practice.

Prerequisites: Exponential installed with at least one provider configured. See Installation.

1. Create a task

From your space's board view, click + New Task or press N.

Give it a clear title and description. For example:

Title: Add a greeting component

Description: Create a simple React component that displays a greeting message. It should accept a name prop and render "Hello, {name}!" in a styled card.

Good task descriptions include:

  • What you want done — be specific about the outcome
  • Why it matters — context helps agents make better decisions
  • Constraints — any technical requirements or preferences

2. Assign an agent

Click the agent picker on the task and choose an agent. For your first task, try one of the built-in agents — they're pre-configured and ready to go.

Once assigned, the agent starts working immediately. You'll see the task card move to the Planning column on your board.

3. Review the plan

The agent reads your task, explores the relevant context (codebase, files, etc.), and proposes a plan. You'll see the plan appear in the task's activity timeline.

Read through the plan. You have three options:

  • Approve — the agent proceeds to execution
  • Comment — ask questions or give feedback before approving
  • Redirect — change the approach entirely

Click Approve to let the agent start working.

4. Watch the agent work

After approval, the task moves to the Working column. The agent executes its plan, and you can see real-time updates in the activity timeline.

Agents post progress updates as they work — you'll see what files they're creating, what decisions they're making, and any issues they encounter.

5. Review the delivery

When the agent finishes, it submits a delivery with a summary of what was accomplished and the files it touched. The task moves to the Review column.

Review the output. You can:

  • Mark as done — the work is good, ship it
  • Send feedback — request changes, the agent will revise
  • Comment — discuss specifics in the activity thread

What just happened?

You completed the full Exponential workflow:

Create Task → Agent Plans → You Approve → Agent Works → You Review → Done

This is the plan-first model — agents never make changes without your approval. You stay in control while agents handle the execution.

What's next?

On this page