Replit
Replit is a browser-based IDE with an autonomous Agent 3 that builds, tests, and deploys apps from natural language, running up to 200 minutes without supervision. Core at $20/month is the real entry point; Starter is a demo tier. Effort-based billing, documented agent reliability issues, and a pattern of billing and support complaints are material risks for production use.
What is Replit?
Replit is a browser-based IDE and hosting platform whose defining feature is Replit Agent, an autonomous AI coding agent embedded directly in the development environment. Unlike code-completion tools such as GitHub Copilot or Cursor, which primarily suggest snippets, Agent handles end-to-end workflows: it plans projects, sets up environments, installs dependencies, writes and edits code across files, connects databases and APIs, tests functionality in a real browser, debugs issues it finds, and deploys live applications. Users interact through plain-English chat prompts, optionally attaching screenshots, files, or URLs. No local installation is required; it works on laptops, mobile apps, and any browser.
The product has evolved through several agent generations. Agent v2 launched in February 2025. Agent 3, released in September 2025, was the major step change: it extended autonomous runtime to 200 minutes or more with minimal supervision, introduced browser-based app testing where the agent drives a headless browser, clicks through the UI, calls APIs, queries the database, and patches regressions it finds, and added the ability to build other agents and automations (Slack bots, Telegram bots, scheduled workflows). Agent 3 runs on Inngest for durability, retries, and observability. Agent 4 followed in March 2026.
A structural characteristic of Replit is that its pricing is inseparable from its agent architecture. Agent usage is billed by effort—request complexity, reasoning steps, tool calls—rather than by message count or token count. This makes light usage relatively cheap and heavy autonomous work potentially expensive in ways that are not predictable from a plan card. The platform's own documentation recommends starting with Free Mode for everyday work and choosing Power mode only when needed, with spend limits configured in advance.
How does Replit work?
A user creates or opens a Replit workspace, selects the Agent tab, and enters a natural-language prompt. The agent analyzes the request, proposes a plan, and begins building. It operates within a multi-language workspace, automatically suggesting a stack based on the prompt (commonly JavaScript/TypeScript with React or Node.js for web apps, Python with Flask or Django for backends, and React Native/Expo for mobile). During the build, Agent 3 periodically opens a headless browser to test the app—checking buttons, forms, APIs, and data sources—and replies with a summary of its tests, fixing any issues it detects. With Max Autonomy enabled, it runs for up to 200 minutes, managing its own task list and monitoring its own progress.
Agent 3 can also build other agents and automations. Selecting “Agents & Automations” from the homepage lets a user describe a Telegram bot, Slack agent, or time-based automation in natural language. These are deployed to an external surface rather than previewed in the workspace, and the Agent provides an admin dashboard for testing. Connections to services like Notion, Linear, Dropbox, and SharePoint are handled through a guided authentication flow rather than manual API-key pasting.
Output is delivered as a running application with a live preview, a deployed URL, generated code files, or a functioning agent deployed to an external platform. The developer reviews the result and iterates through further prompts. Replit's hosting infrastructure—Autoscale, Reserved VM, and Scheduled Deployments—runs the deployed app, and publishing costs are deducted from monthly credits.
Use Cases
- Going from a natural-language description to a deployed full-stack web app without local environment setup
- Building internal tools, prototypes, and MVPs where speed matters more than architectural polish
- Creating Slack bots, Telegram bots, and scheduled automations from a description, without wiring up infrastructure manually
- Letting an agent test and fix its own output in a browser, reducing the manual QA loop for small-to-medium projects
- Running a browser-based development environment on a Chromebook, tablet, or mobile device where a local IDE is impractical
- Using Free Mode for everyday iteration and Power mode selectively for harder tasks, with spend limits configured to bound cost
Pricing
Pricing model: Freemium
| Plan | Price | Billing | Included usage | Notes |
|---|---|---|---|---|
| Starter | $0 | per month | Daily Agent credits up to a monthly cap, limited monthly cloud credits, Lite build only (Full build requires Core or Pro), one published app with a link that expires after 30 days, no Plan mode, no third-party connectors, no Full build. Credits reset daily but are capped monthly. | — |
| Core | $20 | per month | $20 in monthly credits, full Agent with Plan and Build modes, 1 parallel Agent task, up to 5 collaborators, unlimited published apps, 7-day database restore window, 5-hour credit reset with weekly limits. $204/year when billed annually ($18/month equivalent). Core credits do not roll over. | — |
| Pro | $100 | per month | More Free Mode usage than Core, up to 10 parallel Agent tasks, up to 15 builders, tiered monthly credits with volume discounts, 28-day database restore window, unused credits roll over for two months, priority support. $90/month equivalent when billed annually. | — |
| Teams / Enterprise | Custom | custom | Centralized billing, admin controls, SSO, audit logs, and custom credit arrangements. Enterprise adds advanced governance and compliance provisions. | — |
Common Limitations
- Effort-based billing creates unbounded cost risk · Replit Agent bills by effort rather than by message or token. Paid work in Power or Max mode is deducted from credits based on request complexity, and Plan mode can perform billable reasoning even when it produces no code change. Users have documented the agent entering loops—reading files repeatedly or retrying failed fixes—that inflate billable actions, with one report describing effective charges of $50/hour and another describing a single project generating a bill in the high three or low four figures during the Agent 3 release window. Core's $20 monthly credit is exhausted quickly under heavy Power or Max usage, and overage is billed once the allowance runs out. Replit offers spend limits, but the default configuration does not cap the cost of a runaway agent session.
- Agent reliability degrades on larger projects · Multiple independent reviews and Replit's own community forum document reliability problems that compound as project size grows. In September 2026, Replit confirmed a bug where Agent 3 duplicated complete source files during checkpoint or task merge, affecting multiple projects and causing build failures. Reviews note that beyond roughly 45–55 files the browser IDE becomes sluggish, file search slows to seconds, and the AI context window fills to the point of hallucination. Each new Agent chat session is deliberately scoped and does not carry forward the reasoning behind earlier decisions, so multi-session work requires the developer to reconstruct context manually. Replit's own forum advises that large files, hanging install steps, and overlong prompts can cause the Agent to stop responding.
- Free and low tiers are demos, not build environments · The Starter plan restricts users to Lite build, blocks Plan mode and third-party connectors, and provides only one published app whose link expires after 30 days. There is no way to purchase additional Agent credits on Starter. Core at $20/month is the first tier that includes Full build, Plan mode, unlimited published apps, and the ability to buy credit packs, but its $20 credit allowance is consumed quickly by Power or Max mode work. Replit explicitly does not offer a time-limited trial of Core or Pro, so evaluating the paid product requires committing to a subscription.
- Billing and support complaints are a recurring pattern · Across Reddit, G2, Capterra, BBB, and Replit's own community forum, a consistent complaint pattern appears: paid subscribers losing access to deployed Repls due to billing-system errors or unexplained charges, with support response times of 48–96 hours and in some cases 8 days. G2 reviewers describe the billing system as difficult to cancel and refund-resistant. Replit's status page recorded a billing issue affecting database usage on October 15, 2026, resolved after 4 hours 15 minutes. The hosting billing killswitch is described by users as aggressive, cutting off deployments shortly after a missed payment. These are not isolated incidents; they are a documented operational risk for anyone hosting a production workload on Replit.
- Security and compliance gaps for regulated data · Replit holds SOC 2 Type II attestation and offers a GDPR Data Processing Agreement on eligible commercial plans, with every customer—including free-tier users—receiving an isolated GCP project. However, Replit publishes no Business Associate Agreement (BAA) and has no HIPAA offering on any plan, which rules it out for protected health information. A third-party privacy scan detected 64 trackers and browser fingerprinting on replit.com, assigning a privacy grade of E (31/100). Model improvement is on by default for non-Pro users, and Pro users must opt out in Settings → Personalization. Organizations evaluating Replit for sensitive code or regulated data should treat the compliance profile as conditional rather than assumed.
- Browser IDE cannot scale to large codebases · Replit's browser-based IDE is its core architectural choice and its core scalability limit. Reviews consistently report performance degradation beginning around 45–55 files, with slow file search and context-window exhaustion causing the Agent to hallucinate. The platform is optimized for prompt-to-deployment workflows on small-to-medium projects, not for maintaining large existing codebases. Teams planning to grow beyond a prototype should plan for migration to a local IDE or a different hosting arrangement from day one, rather than assuming Replit will scale with the project.
Key Features
- Agent 3 with up to 200 minutes of autonomous runtime and self-supervision
- Browser-based app testing: agent drives a headless browser, tests UI and APIs, and fixes regressions
- Agent can build other agents and automations (Slack, Telegram, scheduled workflows)
- Zero-setup multi-language workspace with automatic stack suggestion
- Full-stack support including React, Node.js, Python, Flask, Django, React Native, and Expo
Replit vs. Lovable, Bolt.new, and v0 by Vercel
This comparison uses official pricing pages, product documentation, and independently published reviews available as of November 2026. It focuses on the structural differences that determine cost predictability, deployment model, and project-size ceiling—not on declaring which tool builds the best app. All four are prompt-first builders, but they occupy different positions: Replit is a full browser IDE with hosting, Lovable and Bolt.new are browser builders with hosted backends, and v0 is a React/Next.js component generator tied to Vercel. The cost and reliability profiles diverge sharply, and those are the dimensions this comparison prioritizes.
| Tool | Positioning | Key differences | Best fit | Pricing note |
|---|---|---|---|---|
| Replit Pricing and plans (official docs) Agent 3 announcement (official blog) Model improvement opt-out (official docs) | A browser-based IDE with an autonomous Agent (Agent 3) that builds, tests, and deploys full-stack apps and automations from natural language, with built-in hosting. | Replit is the only tool in this comparison that is a full browser IDE with built-in hosting across Autoscale, Reserved VM, and Scheduled Deployments. Agent 3 runs up to 200 minutes autonomously and tests its own output in a headless browser. Billing is effort-based rather than credit-per-message or token-per-build, which makes cost less predictable. The free tier is a demo (Lite build only, one expiring app). Documented reliability issues include file duplication bugs and performance degradation around 45–55 files. No HIPAA BAA. | Users who want a zero-setup, full-stack browser IDE with hosting included and an unusually autonomous agent for small-to-medium projects, and who are willing to configure spend limits and monitor billing. | Starter $0 (Lite build, 1 expiring app, daily credits with monthly cap). Core $20/mo ($204/yr) with $20 credits, Full build, Plan mode, 1 parallel Agent. Pro $100/mo with 10 parallel Agents, 15 builders, credit rollover. Credit packs: $100/$100, $300/$290, $500/$480, $1,000/$950. Effort-based Agent billing; overage billed after allowance. |
| Lovable Subscription plans (official docs) Lovable pricing | A browser-based AI app builder focused on full-stack MVPs with authentication and payments, aimed at non-technical founders and fast-moving teams. | Lovable's free plan includes 5 daily build credits (capped at 30/month) plus 20 monthly Cloud credits, and Pro at $25/month provides 100 monthly credits with code editor, custom domains, badge removal, and credit rollover. Lovable does not expose a full IDE in the same sense as Replit—it is a builder with a code editor and Git sync. Its security posture differs: workspace data is excluded from AI training by default on Business and Enterprise, with opt-out per account on Free and Pro, whereas Replit's model improvement is on by default for non-Pro users. Lovable's agent does not have Replit's 200-minute autonomous runtime or browser-based self-testing loop. | Non-technical founders building full-stack MVPs with authentication and payments, who want predictable monthly credit allowances and a lower entry price than Replit's Core tier. | Free: 5 build credits/day (30/month cap), 20 Cloud credits/month. Pro $25/mo ($21/mo annual) with 100 credits, code editor, custom domains, badge removal, rollover. Business $50/mo ($42/mo annual) with governance controls. Enterprise custom. |
| Bolt.new Bolt pricing and tokens (official support) Bolt review (independent, documents training default) | A browser-based AI app builder from StackBlitz, optimized for fast prototypes and hackathon-style demos, running apps in WebContainers. | Bolt's free plan allows 1M tokens per month with a 300K daily cap, and Pro at $25/month raises this to 10M tokens with rollover. Usage is metered in tokens, which makes the unit of consumption more explicit than Replit's effort-based billing, but token consumption still varies with build size. Bolt runs apps in browser WebContainers rather than a hosted cloud IDE, which means native dependencies, GPU access, and long-running background jobs are not supported without moving to other hosting. From October 7, 2026, content on individual plans outside the EEA, UK, and Switzerland can be used for model training unless the user opts out—a broader default training posture than Replit's Pro-only opt-out. | Fast prototypes, hackathon demos, and throwaway apps where token-metered billing and a browser-only runtime are acceptable, and where the project does not need native dependencies or long-running backend processes. | Free: 1M tokens/month, 300K daily cap, Bolt branding, public and private projects. Pro $25/mo with 10M tokens, rollover, custom domains, no branding. Pro+ $50–200/mo with scaled tokens. Teams $30/member/mo. Enterprise custom. |
| v0 by Vercel v0 pricing (official) v0 docs: pricing plans | Vercel's AI builder focused on generating clean React and Next.js components and pages, tightly integrated with Vercel's deployment platform. | v0 is the narrowest of the four: it is optimized for React/Next.js UI generation and Vercel deployment, not full-stack browser IDE work. Its free tier includes $5 in monthly credits, Premium is $20/month, Plus is $30/user/month, and Business is $100/user/month. v0's credit-based pricing has been criticized for exhausting quickly on iterative multi-file workflows, with v0 Max at $5/1M input tokens adding up on complex projects. It does not offer Replit's autonomous 200-minute agent runtime, browser-based self-testing, or built-in hosting beyond Vercel's platform. Its strength is component quality and deployment integration for teams already standardized on Next.js and Vercel. | Front-end developers and teams already using Next.js and Vercel who want high-quality React component generation and one-click deployment without adopting a full browser IDE or autonomous agent. | Free: $5 monthly credits. Premium $20/mo. Plus $30/user/mo with $30 credits. Business $100/user/mo for privacy-conscious teams. Credit-based; overage possible on iterative workflows. |
Comparison checked against official product information on 2026-11-03. Features and pricing may change.
Frequently Asked Questions
Is Replit free?
There is a free Starter plan, but it is a limited trial rather than a usable build environment. Starter restricts you to Lite build, blocks Plan mode and third-party connectors, and gives you one published app whose link expires after 30 days. There is no way to buy additional Agent credits on Starter. Core at $20/month is the practical entry point for Full build, Plan mode, unlimited published apps, and credit top-ups. Replit does not offer a time-limited free trial of Core or Pro.
How does Replit Agent billing actually work?
Agent usage is billed by effort, not by message or token count. Paid work in Power or Max mode is deducted from your credits based on request complexity, and Plan mode can perform billable reasoning even when it produces no code change. Core includes $20 in monthly credits; once those are used, overage is billed or you can buy credit packs ($100 for $100, $300 for $290, $500 for $480, $1,000 for $950). Because effort-based billing does not map to a predictable number of prompts, heavy or looping agent sessions can consume credit much faster than a plan comparison suggests. Replit lets you set spend limits, and you should configure one before starting a large build.
What are Replit's documented reliability issues?
Replit's community forum and independent reviews document several patterns. In September 2026, Replit confirmed a bug where Agent 3 duplicated complete source files during checkpoint or task merge, affecting multiple projects. Reviews report that beyond roughly 45–55 files the browser IDE becomes slow, file search takes seconds, and the context window fills to the point of hallucination. Each new Agent chat session is scoped and does not carry forward earlier reasoning, so multi-session projects require manual context reconstruction. Large files, hanging install steps, and overlong prompts can also cause the Agent to stop responding. These are operational realities to plan around, not isolated incidents.
Does Replit train on my code?
Replit's model improvement is on by default for non-Pro users. Pro users can opt out by going to Settings → Personalization and disabling Model Improvement. The opt-out is Pro-only according to Replit's documentation; lower-tier users do not have a documented self-serve switch. On the compliance side, Replit holds SOC 2 Type II and offers a GDPR DPA on eligible commercial plans, with an isolated GCP project per customer even on the free tier. It does not offer a HIPAA BAA on any plan.
Is Replit suitable for production workloads?
Replit is suitable for prototypes, internal tools, and small-to-medium applications where the build-speed advantage outweighs operational risk. It is not a good fit for production workloads that require high availability. Reviews and forum threads document billing-system errors taking deployed apps offline, support response times of 48–96 hours in some cases, and an aggressive hosting billing killswitch that cuts off deployments shortly after a missed payment. The browser IDE also degrades on larger codebases, with reviewers recommending migration around 45–55 files. If you deploy to Replit, keep a fallback plan and monitor billing closely.
Alternative Tools
-
Bolt.new Bolt.new is StackBlitz's in-browser AI full-stack app builder. It runs a Node.js runtime inside your browser tab, installs packages, starts dev servers, and generates full-stack code from prompts—no local environment setup required. -
Lovable Lovable is an AI full-stack app builder that turns natural-language descriptions into working web applications with frontend, backend, database, auth, and deployment. It is unusually effective at getting to a functional prototype and unusually poor at producing secure production code without human intervention; third-party scans find security issues in 99% of Lovable projects, and the managed backend creates migration friction that the code-portability narrative does not fully offset. -
Cursor Cursor is an AI-first code editor built on a VS Code fork with agent mode, Tab completions, and cloud agents. -
GitHub Copilot GitHub Copilot is an AI coding agent embedded across GitHub, IDEs, the terminal, and the web. It combines unlimited completions on paid plans with agent mode, asynchronous coding agents, and code review, but its value depends on disciplined review, awareness of IP and data-training policies, and realistic expectations about PR acceptance rates and context limits. -
Ollama Ollama is an open-source local LLM runtime and model manager that makes running open-weight models on your own hardware a single command. Local inference is free and unlimited; the costs are hardware, VRAM-gated context defaults, and a documented security surface that requires deliberate network isolation. -
OpenAI Codex OpenAI Codex is an agentic coding product bundled into ChatGPT plans, spanning CLI, IDE, desktop, and cloud. Plus at $20/month is the real entry point; Free and Go are demos. Usage is credit-metered with wide ranges, ChatGPT and Codex pools merge in December 2026, and disclosed security boundary issues remain relevant for teams evaluating it for sensitive codebases.