Replit simplifies Agent modes to Lite, Economy, and Power

What changed

Replit reworked the Agent mode selector in the chat input. The previous dropdown that exposed multiple tuning parameters is now a three-option segmented control:

1. Lite fast, cheap, suitable for small changes and quick questions.
2. Economy balanced; Replit's recommended default for most coding tasks.
3. Power highest capability; use for complex agentic tasks, debugging, or when Economy gets stuck.

Each mode maps to a pre-tuned combination of model, tool budget, and iteration cap. You pick the tier, Replit handles the rest.

Why this matters

The old dropdown surfaced implementation details (model name, iteration count, tool budget) that most users did not want to think about. The new tiered naming matches what users actually want: a rough sense of "how smart and how expensive do you want this to be" without having to understand the plumbing.

Power users keep all the flexibility; the tiers are opinionated presets, not hard-coded limits. You can still override tool access and model selection through the settings panel.

Pricing impact

The mode you pick affects credit burn. Lite is cheapest per turn; Power uses significantly more credits but is more likely to succeed on hard tasks without looping. Replit's rule of thumb in the changelog: start in Economy, drop to Lite for trivial edits, escalate to Power only when Economy cannot get unstuck.

Availability

The new selector is live for all users on web and in the Replit mobile app. No setting to enable; the old dropdown is gone.