Replit Review 2026: Agent Pricing, Export & Hosting

Disclosure: CodeToolPilot may earn a commission if you buy through links on this page, at no extra cost to you. How we make money.

Last verified: October 12, 2026. This is an editorial research review — how we review. Prices and plans change often; confirm on the vendor’s site.

Replit is a cloud development platform that combines a browser-based workspace, managed databases and hosting with Replit Agent, an AI agent that plans, builds, tests and fixes apps from a chat. It started as an online IDE for learners. Today it targets founders, product teams and developers who want one place to build and run software without managing local toolchains or separate hosting accounts.

Underneath the agent sits a full workspace with a shell, Git and package management. For developers, the key questions are how portable your code and data are, and how effort-based pricing turns agent work into a bill. This review covers those questions and the platform’s security tooling.

It is an editorial research review. It is based on Replit’s documentation, pricing page, blog and changelog as of October 2026, plus attributed press coverage. We did not run hands-on tests.

Quick verdict

Replit is the most general-purpose of the major app builders. Agent 4, launched in March 2026, adds parallel tasks, a design canvas and many output types. Under it sits a real workspace: Nix-managed environments, a Git pane, a shell, four deployment types and separate development and production databases. Its security tooling is extensive. The weak spot is cost predictability. Paid agent work is billed by effort, and hosting, databases and some third-party AI calls draw from the same credit balance. Replit does provide usage limits and a hard spending cap.

Pro: A full cloud workspace with a shell, Nix-managed packages and a Git pane with GitHub integration.
Pro: Four deployment types (Autoscale, Static, Reserved VM, Scheduled) with custom domains.
Pro: Separate development and production databases; development databases are free.
Pro: A Security Center with free dependency scans, agent code scans, black-box pen tests and an optional publish block on critical issues.
Con: Effort-based pricing is hard to forecast, and Plan Mode can involve billable reasoning even without code changes.
Con: Hosting, databases, agent work and some third-party AI API calls all draw on the same credits.
Con: The free Starter plan is narrow: Lite builds only, and one published app that goes down after 30 days.
Con: Published apps run on Replit’s own infrastructure, in the US by default.

Visit Replit

What is Replit?

Replit runs each project in a cloud workspace. Its docs say “Replit uses Nix to manage packages and environments,” which is why it can handle many languages and runtimes, not only JavaScript. Replit Agent sits on top. It can set up a project, write code, test its own work and publish the result. Agent 4, announced on March 11, 2026, organizes requests into tasks, runs independent tasks in parallel, and adds a design canvas for comparing layout variants before you commit to code. One project can include web apps, mobile apps, data dashboards, slides and documents that share the same backend.

In 2026 Replit also restructured its plans around Core and Pro, and introduced Free Mode in August 2026. Free Mode, launched with OpenAI, covers everyday agent work at no extra cost for paid subscribers, within an allowance. In October 2026, a changelog entry announced a private preview of Replit Desktop for Windows, which runs builds locally in isolated sandboxes.

Key features

Replit Agent modes, planning and parallel tasks

The agent offers Free, Auto, Power and Max modes. Free covers everyday work within your allowance. Power and Max are paid tiers, and Auto picks a model for paid work. Replit asks you to confirm before paid work starts. Plan Mode breaks work into an ordered task list you can edit before approving it. Background tasks run in parallel: one at a time on Core, up to ten on Pro. Checkpoints let you roll back to an earlier state, and since October 2026 the agent can also send tasks to your other projects from a single chat.

Workspace, Git and export

The Git pane can initialize a repository, connect a remote, stage and commit changes, create and switch branches, and push and pull. Git commands run in the shell stay in sync with the pane, and the gh and glab CLIs are supported. Self-hosted Git providers require Enterprise. You can also export any project with “Download as zip.” Imports work from GitHub, Bitbucket, a zip file, or the GitHub repository behind a Lovable, Bolt or Vercel project. Replit’s docs warn that secrets and existing database data are not migrated during import.

Deployments

Autoscale scales with traffic and bills compute units and requests. Static hosts sites that don’t need a server. Reserved VM keeps an always-on machine with predictable cost. Scheduled runs jobs on a timetable. Every type supports custom domains. Published apps run on Replit’s infrastructure, which its docs say is backed by Google Cloud in the United States; EU hosting is an Enterprise sales conversation.

Databases with dev and production separation

Each app gets a free development database (20 GiB, or 50 GiB on Pro and Enterprise). Production databases are billed on active compute time and storage, with up to 100 GiB on paid plans. This separation became a priority after a widely reported July 2025 incident, in which Replit Agent deleted a user’s production database during a code freeze. Replit’s CEO apologized, pointed to one-click restores from backups, and committed to automatic dev and production separation and a planning-only mode. Pro also extends database rollback to 28 days.

Security Center

Automatic dependency scans are free and cover Node.js, Python, Go, Rust, PHP and Ruby. Paid builders can run an Agent security scan, which audits the codebase, and a Level 3 scan, which adds a black-box penetration test of the running app. A setting can block publishing while critical findings are open, and Enterprise can make it mandatory. Remediation work is billed like other agent work. Build-time checks for insecure patterns and hardcoded secrets are, in Replit’s words, “an early, lightweight check.”

Integrations and teams

The agent can query connected services such as BigQuery, Linear, Slack and Notion, and a Databricks integration became generally available in September 2026. Pro supports up to 15 builders with pooled credits and up to 50 viewers. Enterprise adds SSO/SAML, single-tenant environments and static outbound IPs.

Who should use Replit

Replit suits builders who want an agent and a complete cloud environment in one place: teams building internal tools, data apps and back-end services, and developers working in Python or other non-JavaScript stacks. Small teams that want pooled credits without per-seat fees will find Pro a good fit.

Who should skip it

Skip Replit if you need a fixed monthly cost for AI work, if your compliance rules require hosting on your own cloud account below the Enterprise tier, or if you mainly want an AI assistant inside an existing local repository and IDE.

Strengths and limitations

Strengths

Replit gives the agent a real computer to work on. The Nix-based workspace, shell and Git pane let you drop below the chat whenever you need to, and the four deployment types cover cases most app builders ignore, such as scheduled jobs and always-on bots.

Portability is reasonable: push to your own Git remote or download a zip at any time. On safety, separate development and production databases, checkpoints, the optional publish block and black-box pen testing form a more layered approach than most competitors offer.

Limitations

The billing model is the most common source of frustration. Paid agent work is priced by effort per checkpoint. Replit’s docs say Plan Mode “can perform billable reasoning even when it does not change code.” Some agent features call third-party APIs, such as Claude, ChatGPT and Gemini, at provider rates deducted from your credits. Published apps and production databases draw from the same pool. Use the budget and shutdown limits from the start.

Lock-in sits mostly in the runtime. Your code is portable, but Replit-specific configuration, secrets and production data have to be rebuilt elsewhere, and its docs do not describe one-click deployment to other clouds. Agent autonomy also carries risk. The 2025 database incident is a reminder to keep production credentials out of agent reach and to review changes before they reach production.

Data handling and residency need checking for regulated teams. Advanced privacy controls, single-tenant environments and EU hosting are tied to Enterprise. Read Replit’s privacy policy before putting proprietary code on lower tiers.

Pricing (as of October 2026)

Plan	Price	What you get
Starter	$0	Daily Agent credits up to a monthly cap; monthly cloud credits; Lite build only; one published app with a Replit badge that goes down after 30 days
Core	$20/month, or $18/month billed annually	$20 in credit toward more powerful models; Free Mode for up to 30 hours of chat and 60 projects; Plan Mode; up to 5 collaborators; one background task; 100 GiB outbound data transfer
Pro	$100/month, or $90/month billed annually (larger credit tiers available)	$100 in credit toward the most powerful models; more Free Mode usage; up to 15 builders with pooled credits and 50 viewers; 10 parallel agents; credits roll over for two months; 28-day database rollback; priority support
Enterprise	Custom	SSO/SAML, advanced privacy controls, single-tenant environments, static outbound IPs, custom seat limits

Usage beyond your included credits is billed as you go. Replit’s billing docs describe usage limits, a hard “service shutdown” limit and credit packs with volume discounts. The public pricing page lists Core, Pro and Enterprise, and the billing docs describe the free Starter plan. Confirm current prices, credit tiers and usage rates on Replit’s pricing page.

See Replit pricing

Prices verified on October 12, 2026 against the official Replit pricing page. Plans change often — confirm on the vendor’s page before you buy.

Use cases

Replit is the broadest of the major app builders: an agent on top of a full cloud workspace, with several deployment types and separate development and production databases. These scenarios follow its documented features, and each one notes the caveat that matters most.

Python services, bots and scheduled jobs

Developers working outside JavaScript can use Replit because it manages packages and environments with Nix, so the same workspace handles Python and other runtimes. Reserved VM deployments keep an always-on machine for bots or workers, and Scheduled deployments run jobs on a timetable. Published apps run on Replit’s infrastructure, which is backed by Google Cloud in the United States, and EU hosting is an Enterprise sales conversation.

Taking over a project started in another app builder

A developer can import the GitHub repository behind a Lovable, Bolt or Vercel project, or any GitHub, Bitbucket or zip source, and continue with Replit’s shell, Git pane and agent. In our assessment, this suits teams that have outgrown a JavaScript-focused builder and want more control over the runtime. Replit’s docs warn that secrets and existing database data are not migrated during import, so plan to recreate them.

Internal data apps for a small team

The agent can query connected services such as BigQuery, Databricks, Linear, Slack and Notion, which makes internal dashboards and reporting tools a natural fit. Pro supports up to 15 builders sharing pooled credits and up to 50 viewers. Production databases, hosting and agent work all draw from the same credits, so set usage limits before the tool becomes popular.

Supervised agent work on a live app

Teams can let the agent change a production app while keeping guardrails: separate development and production databases, checkpoints for rollback, Plan Mode to approve a task list first, and an optional block on publishing while critical security findings are open. Pro extends database rollback to 28 days. Plan Mode can involve billable reasoning even when it does not change code, and the 2025 incident in which Replit Agent deleted a user’s production database is a reminder to keep production credentials out of the agent’s reach.

Parallel feature work across projects

On Pro, up to ten background tasks can run in parallel, and since October 2026 the agent can send tasks to your other projects from a single chat. A team lead can queue independent fixes across several apps and review them as they finish. Core allows only one background task, and effort-based billing means parallel work can use credits quickly.

Alternatives

Lovable: a focused web app builder with a built-in Supabase-based backend and Git sync on every plan.
Bolt.new: a fast, browser-based JavaScript app builder with two-way GitHub sync and built-in hosting.
v0: a Next.js-first builder with branch-per-chat pull requests and Vercel deployments.
Claude Code: a terminal agent for developers who want agentic coding against their own repository and infrastructure.

Final verdict

Replit is the most capable all-in-one option in this category, with a genuine workspace, flexible deployments, separate dev and production data, and strong security tooling around its agent. It rewards developers who want to work below the chat when needed. Go in with budgets configured, because effort-based pricing and shared credits make monthly costs harder to predict than the headline plan prices suggest.

Try Replit

FAQ

Is Replit free?

There is a free Starter plan, but it is narrow. It includes daily Agent credits up to a monthly cap and monthly cloud credits, limits you to Lite builds, and allows one published app with a Replit badge that goes down after 30 days. Free Mode, launched with OpenAI in August 2026, covers everyday agent work at no extra cost for paid subscribers, within an allowance.

Can I export my code from Replit?

Yes. The Git pane can push to your own remote, the gh and glab CLIs work in the shell, and any project can be exported with “Download as zip.” Replit-specific configuration, secrets and production data have to be rebuilt elsewhere, and the docs do not describe one-click deployment to other clouds. Self-hosted Git providers require Enterprise.

What languages does Replit support?

Replit manages packages and environments with Nix, so it supports many languages and runtimes rather than only JavaScript. Its free automatic dependency scans cover Node.js, Python, Go, Rust, PHP and Ruby. In our assessment, that makes Replit the most flexible of the major app builders for Python or other non-JavaScript back ends.

Does Replit use my code to train AI models?

It depends on your plan. Replit’s model-improvement page says Pro users can opt out by turning off Model Improvement under Settings, Personalization, that the opt-out applies to new content, and that Replit does not use Enterprise customers’ content to train its models. The page does not describe an opt-out for Starter or Core, so read Replit’s privacy policy before putting proprietary code on those plans.

Can Replit Agent delete my production database?

It has happened: in July 2025, Replit Agent deleted a user’s production database during a code freeze, and Replit’s CEO apologized and committed to automatic separation of development and production databases. Today each app gets a separate development database, checkpoints let you roll back, and Pro extends database rollback to 28 days. As guidance, keep production credentials out of the agent’s reach and review changes before they reach production.

Can I run Replit locally or host apps on my own cloud?

Only in limited ways for now. In October 2026, Replit announced a private preview of Replit Desktop for Windows, which runs builds locally in isolated sandboxes. Published apps run on Replit’s infrastructure, backed by Google Cloud in the US, and single-tenant environments and EU hosting are tied to Enterprise. If you must host on your own cloud account, export the code and deploy it yourself.

Replit Review (2026): Is Replit Agent Worth It for Developers?