Replit review, pricing and verdict

Last updated: 2026-11-02

Replit is a browser-based platform where Agent 4 turns a plain-English description into a working, hosted app with its own database and login. Since August 2026 a Free Mode handles everyday edits without spending credits, and since October a Monitoring Agent can find and fix production errors on its own.

About Replit

Replit is a browser-based building platform, founded in 2016 by Amjad Masad, where you describe an app, a website, a slide deck or a data dashboard in plain English and its AI agent writes the code, sets up the database and publishes it. The agent is now in its fourth generation, Agent 4, and nothing has to be installed on your computer.

The biggest change this year is how AI usage is paid for. Since mid-August 2026 a Free Mode, running on OpenAI's GPT-5.6 Luna, handles everyday chat and edits without using credits on paid plans, within limits that reset every five hours. Heavier jobs move to Power Mode or Max Mode, which draw on credits under effort-based pricing, and Replit asks before a paid action starts. Since late August, intelligent model routing picks the model for each task automatically. Projects can include web and mobile apps, slides, videos and charts, and the agent tests its own work and saves checkpoints you can roll back to.

October 2026 pushed Replit further into company use. A Monitoring Agent watches published apps, investigates production errors and can ship a validated fix; Cloud Agents let non-engineers propose changes to a company's own codebase for engineers to approve; and Enterprise customers can now run Replit in their own cloud account. A native Databricks integration, generally available since September, lets internal apps read governed company data.

It works well for startup founders, PMs and lean teams that need a working prototype or internal tool without a separate host, database or deployment pipeline, and developers who want to start a project from any browser. Our guide on building AI apps without being a developer walks through that path, and the AI app builder roundup ranks it against its peers. Plan prices sit in the pricing notes below.

Replit is not for everyone. Engineers with an existing repository who want an AI pair programmer in their editor usually prefer Cursor or a terminal agent such as Claude Code. Founders who only want a polished React front end often shortlist Lovable, while the similar Bolt.new is another quick prompt-to-site option, and v0 is lighter if all you need is interface components. Paid work outside Free Mode is metered, so a long session in Max Mode can cost far more than the subscription.

Pricing

Starter is free: daily Agent credits up to a monthly cap, Lite build only, and one published app that comes down after 30 days. Core costs $20 a month, or $18 a month billed yearly, and lets you chat in Free Mode for as much as 30 hours a month, plus $20 a month towards stronger models, 5 collaboration seats and unlimited published apps. Pro costs $100 a month, or $90 billed yearly, with $100 of model credit, 15 collaborators, 28-day database rollback and premium support; larger yearly Pro credit levels run $215, $425, $825 and $2,000 a month, and Pro credits roll over for two months.

Enterprise is quoted by sales and adds SSO, SCIM, audit logs and single-tenant networking. Beyond the included credit, Power and Max Mode work is billed by effort, and app hosting, databases and storage draw on the same credits. For a lighter front-end builder, Emergent or 10Web may cost less for a simple site.

Key Features

• Agent 4: Replit's current agent plans, writes, tests and checkpoints a project from a plain-English brief, with Kanban task planning and background tasks on paid plans.
• Free Mode: Launched in August 2026 on OpenAI's GPT-5.6 Luna, it covers everyday Agent work without spending credits on Core and Pro, within limits that reset every five hours.
• Power and Max modes: Two paid modes for heavier work, billed by effort, with Replit asking for confirmation before a paid action begins.
• Monitoring Agent: Released 22 October 2026, it detects production issues in a published app, investigates the cause, validates a fix and can publish it.
• Cloud Agents for company code: Since 27 October 2026, people outside engineering can request changes to a company's own repository, with engineers approving what ships.
• Built-in infrastructure: Authentication, databases, object storage and hosting come set up, and every customer gets an isolated Google Cloud project, even on the free tier.
• Databricks integration: Generally available since 10 September 2026 with Lakebase support, letting apps work on governed warehouse data with automatic preview deploys.

Pros

• Free Mode removes most of the meter anxiety for everyday edits, which was the biggest complaint about Replit's earlier credit model.
• Hosting, database, authentication and publishing are already wired together, so a prototype can go live without any infrastructure work.
• Per-customer cloud isolation and pre-publish security scanning are stronger defaults than many prompt-to-app builders offer.
• One workspace covers apps, slides, charts and video, so a small team does not need a separate tool for each output.
• The Monitoring Agent and Cloud Agents extend Replit from prototyping into maintaining real company software.

Cons

• Work outside Free Mode is billed by effort, so a long Max Mode session can cost several times the monthly subscription.
• Free Mode limits reset every five hours and also carry weekly caps, which can interrupt a long build day.
• The free Starter plan only keeps one published app online for a month, so it is a trial rather than a hosting option.
• ISO 27001 certification is still in progress, according to Replit's security page, which matters for some procurement checklists.
• Everything runs in the browser and needs a live connection; there is no offline mode.

Visit Replit Official Website