# Replit Agent — Chat-driven agent that builds, tests and publishes full web and mobile apps on Replit's cloud.

> Source: The Agents Index — https://theagentsindex.com/replit-agent (structured, researched, re-verified)
> Facts last verified: 2026-09-22

Replit Agent is a cloud coding agent that turns a plain-language description into a working web or mobile application inside Replit's hosted workspace. It writes the code, tests it in a browser, and publishes the result to Replit's infrastructure.

| Fact | Value |
| --- | --- |
| Website | https://replit.com/products/agent |
| Price | from $20/month billed monthly, $18/month billed annually |
| Free tier | Yes |

Starter plan: daily Agent credits up to a monthly cap, monthly cloud credits, Lite build only (Full build needs Core), 1 published app that goes down after 30 days, no background tasks, no third-party connectors, no Plan Mode, 'Made with Replit' badge retained.

## What it does

**Writes code** · Yes — Describing an app in chat is the entire entry path; Agent authors the project from that prompt, no prior code needed.

> Tell Replit Agent your app or website idea, and it builds it through chat. It's like having an engineering team on demand, with no coding experience required.
> — https://replit.com/products/agent · Hero section, beneath the page headline

**Debugs and fixes** · Yes — Agent runs a reflection loop that finds problems in the running app and lands the fix itself, rather than only reporting them.

> Agent tests and fixes its code, constantly improving your application behind the scenes in a reflection loop.
> — https://replit.com/products/agent · Feature section 'Agent tests its own work so you dont have to'

**Writes tests** · Partly — The vendor describes browser-driven testing of the running app and a generated report, not authored test files committed to the repository.

> While building your app, Agent will now periodically test it using a browser, generating a report and fixing the issues it finds.
> — https://replit.com/products/agent · Feature section 'Agent tests its own work so you dont have to'

**Reviews code** · Partly — Documented review is Agent checking its own output as it builds; no documented flow for reviewing a diff or pull request a person wrote.

> Agent reviews its own code automatically as it builds, catching mistakes and reducing costly rework—there's no setting to turn on or off.
> — https://docs.replit.com/billing/managing-spend · Section 'Control costs with Agent modes'

**Whole-codebase work** · Partly — Max Mode is positioned for larger codebases and tasks can carry dependencies, but no repository-wide migration guarantee or size limit is published.

> Max Mode uses more capable models for complex tasks, larger codebases, and harder problems.
> — https://docs.replit.com/features/agent/agent-modes · Section 'Max Mode'

**Ships it** · Yes — Agent does not stop at a diff: publishing puts a running instance on Replit's cloud, with custom domains and monitoring included.

> When you publish your Replit App, you create a published app. A published app is a running instance of your app on Replit's cloud infrastructure.
> — https://docs.replit.com/learn/projects-and-artifacts/replit-deployments · Section 'What is Publishing?'

**Done-for-you service** · No — This is self-serve software bought by subscription; Enterprise adds named advocates for adoption help, not a team that builds your app.

> Dedicated Product Advocate and Field Engineer: Named Replit partners who help your teams adopt Replit, scale up, and unblock.
> — https://docs.replit.com/billing/plans/replit-enterprise · Key features list

## How much it does unattended

**Runs autonomously** · Partly — Background tasks run unattended in isolated copies, but nothing reaches your main version until you review and apply each result.

> Background tasks are separate threads where Agent works independently in isolated copies of your project. Background tasks do not change your main version until you review and apply them.
> — https://docs.replit.com/core-concepts/agent/task-system · Section 'What is the task system?'

**Multi-agent** · Yes — Several tasks run in parallel with automatic dependency detection and AI-assisted conflict resolution, though the ceiling varies sharply by plan.

> Starter does not support background tasks. Core runs 1 at a time, Pro supports up to 10, and Enterprise supports up to 64.
> — https://docs.replit.com/core-concepts/agent/task-system · Step 'Tasks start running'

**Agent permissions** · Partly — Agent edits directly in Build mode, but paid escalation needs explicit confirmation and Plan Mode gates scope before any change is made.

> Replit asks for confirmation before a paid action starts, and Agent returns to Free Mode afterward.
> — https://docs.replit.com/help/agent-and-ai · FAQ 'What are the Agent modes, and how do they differ?'

## Which models it runs on

**Claude** · Yes — Anthropic models are selectable in both paid modes, with the heavier variants reserved for Max Mode rather than the cheaper Power tier.

> Claude Sonnet 4.6 | ✓ | ... Claude Opus 5 | | ✓
> — https://docs.replit.com/features/agent/model-selector · Table 'Models by mode'

**GPT** · Yes — OpenAI GPT variants appear in the published model catalogue for both paid modes; Free Mode routes automatically and offers no picker.

> GPT-5.6 Luna Fast | ✓ | ... GPT-5.6 Sol | | ✓
> — https://docs.replit.com/features/agent/model-selector · Table 'Models by mode'

**Open models** · Partly — Open-source model providers are a real but second-class category for Agent: they exist as a provider policy and are off unless an admin turns them on.

> Open-source model providers are disabled by default. To make one available, select Enable all or Enable selected, then select Update.
> — https://docs.replit.com/features/collaboration/enterprise-model-controls · Sentence closing the 'Set a provider policy' section

**Model choice** · Partly — Manual model choice exists only in the paid modes on paid plans; Free Mode and Starter always defer to Replit's router.

> Free Mode always uses intelligent model routing. Builders on Core and Pro can choose an available model in Power Mode or Max Mode.
> — https://docs.replit.com/features/agent/model-selector · Section 'Intelligent model routing' note and 'Choose a model manually'

## Where you use it

**In your editor** · Partly — Agent lives in Replit's own browser-based Project Editor; no plugin for a local editor such as VS Code is documented.

> Access version control in your Replit App by adding the Git tool to the Project Editor
> — https://docs.replit.com/learn/projects-and-artifacts/version-control · Section 'Getting started'

**On the command line** · Partly — A full Git command line is available through the workspace Shell, and terminal MCP clients can drive Agent, but there is no local Replit CLI.

> Git CLI offers command-line access to all Git functionality
> — https://docs.replit.com/learn/projects-and-artifacts/version-control · Section 'How version control works on Replit', numbered list

**In your pipeline** · No — The GitHub connector is scoped to repositories, users and organizations; Replit's own connector list gives merge requests and pipelines to GitLab, not to Agent.

> GitHub: Access repositories, users, and organizations
> — https://docs.replit.com/features/integrations/overview · 'Available connectors' section, 'Developer tools' list

**In a browser** · Yes — The entire product is a hosted web workspace: you sign up, chat, build and publish without installing anything locally.

> Tell Agent what you want to build, refine it through conversation, and launch a production-ready app from one workspace.
> — https://replit.com/products/agent · Closing call-to-action section

## Whose machine it runs on

**Self-hosted** · No — Everything runs in Replit-operated Google Cloud projects; isolation is a dedicated single-tenant GCP project, not infrastructure you own.

> Replit hosts data primarily in Google Cloud Platform (GCP) data centers in the United States, with an optional hosting region in India for users who opt in.
> — https://docs.replit.com/teams/information-security/overview · Section 'Hosting and infrastructure'

_Not established: Open source_

## What it costs to run

**How it meters** · Yes — Billing is a subscription plus effort-based credit consumption per completed checkpoint, with third-party model calls passed through at provider rates.

> Paid modes and confirmed paid work use effort-based pricing based on request complexity.
> — https://docs.replit.com/billing/ai-billing · Section 'Agent billing'

**Free tier** · Yes — The Starter plan is a standing free tier with daily Agent credits, no sales call, though Full build and connectors need Core.

> Starter plan users receive daily credits for Agent usage, up to a monthly cap. These credits reset daily, giving you a fresh allowance to work with Agent each day.
> — https://docs.replit.com/billing/plans/starter-plan · Section 'Daily credits for Agent'

**API access** · Partly — The published REST API is an Enterprise-only admin beta for usage and membership data; it does not drive Agent builds.

> The Admin API is available only to account admins on Enterprise accounts. Account admins can create and manage API keys.
> — https://docs.replit.com/teams/admin-api · Section 'Who can access the Admin API'

**MCP server** · Yes — Replit runs its own MCP server over streamable HTTP with OAuth, exposing tools that create, update and publish apps from any client.

> Replit MCP uses the Model Context Protocol to let you create, find, inspect, update, and publish Replit Apps from AI tools you already use.
> — https://docs.replit.com/platforms/mcp-server · Opening paragraph, Replit MCP Server

**Bring your own key** · Partly — You can supply provider keys for models your own app calls, but the inference behind Agent itself is fixed to Replit's infrastructure.

> You can bring your own AI provider keys for the models your app calls. The model that powers the Agent experience itself runs on Replit's infrastructure and can't be swapped.
> — https://docs.replit.com/help/agent-and-ai · FAQ 'Can I use my own API keys with Agent?'

## Buying it for a team

**A company can buy it** · Yes — Organisations buy Pro or Enterprise directly; Enterprise can be self-served from inside the product without a sales conversation.

> You can upgrade to Enterprise directly inside Replit. Builders on any plan — Free, Core, or Pro — can move to Enterprise without leaving the product.
> — https://docs.replit.com/billing/plans/replit-enterprise · Introduction paragraph

**Seat model** · Partly — Pro charges no per-seat fee but caps builders at fifteen; Enterprise lifts the cap, so the ceiling moves with the plan rather than the headcount.

> Pro has no per-seat fees: you get up to 15 builders with pooled credits and budget controls.
> — https://docs.replit.com/billing/teams-billing/overview · Section 'Replit Pro'

**Pooled budget** · Yes — Team credits sit in one shared balance rather than per-person wallets, with usage tracked per workspace for attribution.

> Pro has no per-seat fees: you get up to 15 builders with pooled credits and budget controls. Everyone works from the same plan.
> — https://docs.replit.com/billing/teams-billing/overview · Section 'Replit Pro'

**Admin controls** · Partly — Real constraint on what builders and Agent may do — private-only publishing, banned exports, pinned geography — is an Enterprise settings pane.

> Ban source code export: Prevents builders from exporting app source code as a ZIP file.
> — https://docs.replit.com/teams/privacy-and-deployment-settings · Section 'Restrict exports and uploads'

**Audit log** · Partly — Audit logging, SIEM streaming and the prompt-retrieval compliance API exist, but only on Enterprise and only for account admins.

> Audit logs are available exclusively for Enterprise customers. Account admins can enable them in Settings.
> — https://docs.replit.com/teams/identity-and-access-management/audit-logs · Introduction note

**Single sign-on** · Partly — SAML SSO with SCIM provisioning is an Enterprise-tier feature; Core and Pro buyers get no federated identity.

> SSO and SAML with SCIM provisioning: Secure, automated team access tied to your identity provider.
> — https://docs.replit.com/billing/plans/replit-enterprise · Key features list

## What happens to your code

**Opt out of training** · Partly — Content trains Replit's models by default; the opt-out switch is documented for Pro, while Enterprise content is excluded outright.

> This control is available to Pro users. For Enterprise users, Replit does not use customer content to train its models.
> — https://docs.replit.com/legal-and-security-info/model-improvement · Closing paragraphs of Model improvement

**Data residency** · Partly — Data sits in US Google Cloud by default with an opt-in India region; EU hosting of published apps requires an Enterprise sales request.

> All published apps are hosted in the United States. Enterprise customers can contact sales to request their published apps to be hosted in the European Union instead.
> — https://docs.replit.com/learn/projects-and-artifacts/replit-deployments · Info callout under 'What is Publishing?'

**Getting out** · Partly — Source leaves cleanly as a zip or through GitHub sync, but purchased credits expire six months out and do not roll over.

> Expiration: Credit packs expire six months after the purchase date ... No rollover: Unused credits from a pack do not roll over after expiration
> — https://docs.replit.com/billing/managing-spend · Section 'How credit packs work'

**Certifications** · Partly — Replit states its own SOC 2 Type 2 attestation; the ISO 27001 claim on the same page belongs to Google Cloud, not to Replit.

> GCP is an industry-leading cloud provider, certified for compliance with ISO 27001 and SOC 2 Type 2. Additionally, Replit has achieved SOC 2 Type 2 Attestation of Compliance
> — https://docs.replit.com/teams/information-security/overview · Section 'Hosting and infrastructure'

## Provenance

Every fact above comes from the vendor. Nothing here is independently corroborated yet.

## Coverage elsewhere

Pages about this product that the vendor does not control. The label states how independent each one is.

| Source | Kind | Read |
| --- | --- | --- |
| techcrunch.com | Independent source | 2026-08-24 |
| the-decoder.com | Independent source | 2026-08-24 |
| inc.com | Independent source | 2026-08-24 |

## What changed

- 2026-09-22 — Capabilities · Commercial terms · Free tier · Price
- 2026-09-02 — Capabilities · Commercial terms · Free tier
- 2026-08-31 — Capabilities · Commercial terms · Free tier · Price
- 2026-08-17 — Pricing tiers.enterprise.features
- 2026-08-16 — One-liner · Summary

## Correcting this record

Anything marked "Not established" is a question we have not answered — it is not a No. If you represent Replit Agent, or anything here is wrong or out of date, you can supply it with a source: the operations this site accepts from an agent are listed at https://theagentsindex.com/.well-known/agents.json — that catalogue needs no key to read, and each entry states its own auth.

Supplying evidence is free. Nothing that can be bought here changes what a field says, whether a gap is closed, or how this record is scored.