Replit

Replit is a full-stack AI coding platform from Replit. The headline feature in 2026 is Agent 3, an autonomous coding agent that runs up to 200 minutes per session, spawns subagents for parallel work, tests its own code, and ships the finished app to a live deployment URL. The free tier is enough to evaluate, and $20-95/month unlocks heavier credit usage and the most powerful underlying models.

What Replit does:

• •Agent 3 autonomous coding describe an app in plain English, and Agent 3 plans, writes, tests, and self-corrects code for up to 200 uninterrupted minutes per session, choosing Economy, Power, or Turbo mode based on the task.
• •Browser-based full IDE a complete cloud IDE running on Replit infrastructure means you write, debug, and run code from any laptop or Chromebook with zero local setup.
• •One-click deployment every project gets a public URL plus optional private password-protected deployments, custom domains, and static outbound IPs on higher tiers.
• •Built-in Postgres and key-value DB ReplitDB plus Postgres on Pro provide persistent storage with a 28-day restore window without leaving the workspace.
• •MCP support Replit ships an MCP server, so Claude Desktop, Cursor, and any MCP client can read or modify your Replit workspaces as a tool call.
• •Real-time multiplayer collaboration 5 collaborators on Core and 15 on Pro can edit, run, and debug the same Repl together with cursors visible per user.
• •Mobile apps for iOS (4.7) and Android (4.6) ship and edit from a phone, with the iOS and Android apps consistently rated above 4.5.
• •Effort-based credits system (Feb 2026 reset) the new pricing ties Agent compute to credit packs that scale with model power, so cheap tasks stay cheap and Turbo mode is reserved for hard work.
• •AI integrations gallery and CLI one-click connections to OpenAI, Anthropic, Stability, ElevenLabs, and others with hosted API keys; the Replit CLI lets you push, pull, and run Repls from any local terminal.
• •Templates and bounties marketplace thousands of starter templates plus a paid bounties board for getting custom Repls built on demand.

Pricing:

• •Starter $0: free Agent credits each day, 1 published app, public Repls only. Best for trying Agent 3 and learning.
• •Core $20/month (billed annually) or $25/month: $25 of monthly credits, 5 collaborators, unlimited workspaces, private deployments. Best for solo builders shipping side projects.
• •Pro $95/month (billed annually) or $100/month: $100 of monthly credits, 15 collaborators, 50 viewers, Turbo mode access to top-tier models. Best for serious indie or small-team work.
• •Enterprise custom pricing: SSO, SAML, single-tenant environments, region selection, VPC peering, data warehouse connections.

Limitations:

• •Credit costs scale fast on Agent after the Feb 2026 pricing reset, builds that used to cost $0.50 of credit can hit $3+ on the first prompt; users report 3-5x overall increases on Agent-heavy workflows.
• •Free tier output is restricted Starter caps you at 1 published app and public-only Repls, so any commercial or private prototype forces an upgrade quickly.
• •Code quality varies by mode Economy mode produces the cheapest output but also the most rework; serious builds need Power or Turbo, which burn credits much faster.
• •No traditional REST API Replit does not expose a public coding API the way GitHub or GitLab do, so deeper automation has to go through MCP or browser-driven workflows instead of curl.