Skip to main content
The chat panel is the main control surface in Openv2. You describe what you want; the AI plans, edits your files, and runs commands — all inside your browser. The result shows up in the Preview and the editor without any manual steps from you.

What happens when you send a message

Every message goes through the following sequence:
1

Credits are checked

Openv2 confirms you have credits available before the turn starts. If you’re out of credits, you’ll see a notice prompting you to upgrade or wait for a daily reset.
2

Your project is shared with the model

A snapshot of your current project files is sent to the AI so it sees your code as it actually exists — not a stale version.
3

The model responds

The AI produces a response — text, a plan, or a mix — and may issue tool calls to read files, write files, edit code, search, or run shell commands.
4

Tool calls run in your WebContainer

File edits and shell commands (like npm install or npx expo ...) are applied directly in your browser’s WebContainer. Your node_modules, dev server, and Preview all stay in sync.
5

The turn is saved as a version

When the turn completes and files changed, Openv2 saves your chat history and creates a new version snapshot — so you can rewind to this point later.

Choosing a model

Use the model selector in the header to pick which LLM backs your session. The available models depend on your Openv2 plan and provider availability.
  • Stronger models are better for architecture decisions, complex refactors, and multi-file changes.
  • Faster models are often enough for small edits, copy changes, and style tweaks.
If a large refactor goes wrong, switch to a stronger model and ask the AI to explain its plan before writing any code.

Credits

Openv2 uses a credit system to meter AI usage. Each completed assistant turn typically uses one credit.
PlanCredits
FreeDaily allowance (resets each day)
ProLarger monthly credit pool via Stripe
UltraHighest monthly credit tiers via Stripe
When your credits run out, Openv2 will prompt you to upgrade or wait for the next daily reset. See Billing & plans for current pricing and limits.

Suggestions and planning blocks

Depending on the assistant mode, the chat UI may show:
  • Suggested follow-up prompts — quick actions based on where you are in the build.
  • Thinking / planning blocks — a summary of what the AI intends to do before it writes any code. Review these to confirm direction before large changes.
  • Implementation plans — structured step-by-step breakdowns for bigger features.
Use planning blocks as a checkpoint. If the plan doesn’t match what you want, correct it before the AI starts writing — it’s much cheaper than reverting after the fact.

Viewer role

If you were invited to a project as a viewer, you can browse the workspace but you cannot use the AI to edit the project. You will see a notice in the chat panel explaining this restriction. To request edit access, ask the project owner to change your role.

Prompting

Write prompts that get better results with fewer turns.

Versions & history

Browse and restore earlier snapshots of your project.

Workbench & editor

Edit files manually and monitor the terminal alongside chat.

Billing & plans

Understand your daily and monthly credit limits.