Pre.dev

Long horizon coding agent that plans, builds and verifies software

Intermediate API
Screenshot of Pre.dev, Long horizon coding agent that plans, builds and verifies software

What is Pre.dev?

Pre.dev is a long horizon coding agent that plans architecture before writing code, builds, verifies and ships into GitHub. It connects to 25 services, holds project conventions as agent skills, supports MCP servers and ships apps with AI, payments and authentication preconfigured.

Pre.dev is a coding agent built around a single idea: describe the software once and let the agent carry the work through a full cycle rather than prompting it line by line. The homepage splits that cycle into five visible stages. The agent takes a natural language description, drafts architecture and a roadmap before any code is written, then builds and shows its work, verifies the result against checks, and ships it into your GitHub with a live preview. The planning stage is the distinguishing piece: work is decomposed into a dependency graph first, which the agent then walks through on long runs. In practice, the agent connects to tools you already run through one OAuth per service. The homepage lists 25 services across 9 categories, including GitHub, Slack, Notion, Linear, Figma, Vercel, Supabase, Sentry, Jira and Salesforce. Beyond that, standing agent skills (project conventions such as TypeScript only, Tailwind only, or a required API style) are injected into every conversation, and MCP servers can be attached over stdio, HTTP or SSE with whole tool sets exposed. API keys for 195 providers across 22 categories are stored encrypted and injected at build time, so your own billing runs through them. Apps built on Pre.dev start with AI, payments and authentication already wired: models from major labs, checkout and subscriptions with payouts, and email plus Google sign in on a Clerk app you can take over. The vendor publishes Terminal-Bench 2.0 results claiming a full model tier down still outscores Claude Code and OpenCode at roughly a third of the cost, and the agent is model agnostic, so you can pick providers per job. It suits founders shipping MVPs, small engineering teams that want reviewed pull requests instead of live pair prompting, and developers pointing the agent at an existing repository to extend it in its own patterns. Compared with chat style assistants, the trade off is less turn by turn control; compared with fully autonomous coding agents, the planning and verification gates are its main claim to difference.

How do you use Pre.dev?

  1. 1Describe the software you want built
    Open the builder and type a description of the app or change. You can import context and pick the underlying model from the selector before starting.
    Pre.dev — Describe the software you want built
  2. 2Review the plan before code is written
    The agent returns architecture and a roadmap decomposed into a dependency graph. Check that decomposition, since the rest of the run follows it.
  3. 3Connect your tools and conventions
    Authorize services such as GitHub, Linear or Supabase once each, and add agent skills describing your stack rules so every conversation inherits them.
  4. 4Attach API keys and MCP servers
    Add keys for the providers your app calls so billing runs through your accounts, and register any MCP servers you need with their transport type.
  5. 5Let it build, verify and ship
    The agent writes code, shows its work and will not close a task until checks pass, then lands changes in your GitHub repository with a live preview.
  6. 6Inspect reference apps before committing
    Three apps built from single prompts are live on the site (Restyle, CandleSeat, Northfolio), which is the quickest way to judge output quality.
    Pre.dev — Inspect reference apps before committing

Pros and cons

Pros

  • Plans architecture and a dependency graph before writing any code, which gives long runs a structure to followAI
  • Verification gates (type checks, lint, browser checks) must pass before work is marked doneAI
  • Model agnostic: run different providers per job instead of being tied to one vendorAI
  • One OAuth per service across 25 integrations, plus MCP servers over stdio, HTTP or SSEAI
  • Apps start with AI, payments and authentication already working, cutting setup work on new buildsAI

Cons

  • Free tier credits are limited and published apps are only live for 7 days, so real testing needs a paid planAI
  • Model selection, custom domains and permanent hosting sit behind Premium at $41 per monthAI
  • The hands off model reduces fine grained control over how individual files are writtenAI
  • Benchmark claims come from the vendor's own published Terminal-Bench 2.0 run, with no independent verification on the pageAI
  • No dedicated support documentation is described on the homepage, so issues are likely resolved through generic channelsAI

How much does Pre.dev cost?

Free trial

A Free tier exists at $0 with 20 credits, browser and terminal coding agent access, and app publishing for 7 days. No separate time limited trial is described.

Learn more

Pricing

Six tiers: Free at $0 with 20 credits; Plus at $8 per month ($99 annually) with 100 monthly credits; Premium at $41 per month ($499 annually) with 500 credits, model selection, permanent apps and custom domains; Pro at $166 per month ($1,999 annually) with 2,500 credits; Team at $415 per month ($4,990 annually) with unlimited seats and 5,000 pooled credits; Enterprise with custom pricing.

Learn more

Integrations

25 services across 9 categories via one OAuth each, including GitHub, GitLab, Bitbucket, Slack, Notion, Linear, Jira, Figma, Vercel, Supabase, Sentry, Airtable, HubSpot, Asana, Salesforce, Zoom and Dropbox. API keys cover 195 providers, and MCP servers connect over stdio, HTTP or SSE.

Features

Five stage build cycle (describe, plan, build, verify, ship) with architecture and a dependency graph up front. Agent skills inject standing conventions into every conversation. 25 service connections via single OAuth, encrypted API keys for 195 providers, and MCP servers over stdio, HTTP or SSE. Built in AI, payments and authentication on every app, plus GitHub delivery with live previews.

Frequently asked questions about Pre.dev

  • How much does Pre.dev cost?
    Six tiers: Free at $0 with 20 credits; Plus at $8 per month ($99 annually) with 100 monthly credits; Premium at $41 per month ($499 annually) with 500 credits, model selection, permanent apps and custom domains; Pro at $166 per month ($1,999 annually) with 2,500 credits; Team at $415 per month ($4,990 annually) with unlimited seats and 5,000 pooled credits; Enterprise with custom pricing.
  • Does Pre.dev offer a free trial?
    A Free tier exists at $0 with 20 credits, browser and terminal coding agent access, and app publishing for 7 days. No separate time limited trial is described.
  • Does Pre.dev have an API?
    Yes, Pre.dev offers an API.
  • How do you use Pre.dev?
    The walkthrough on this page covers 6 steps: 1. Describe the software you want built 2. Review the plan before code is written 3. Connect your tools and conventions 4. Attach API keys and MCP servers 5. Let it build, verify and ship 6. Inspect reference apps before committing.
  • What platforms does Pre.dev support?
    Pre.dev is available on Web App, MacOS, Windows and Linux.
  • What does Pre.dev integrate with?
    25 services across 9 categories via one OAuth each, including GitHub, GitLab, Bitbucket, Slack, Notion, Linear, Jira, Figma, Vercel, Supabase, Sentry, Airtable, HubSpot, Asana, Salesforce, Zoom and Dropbox. API keys cover 195 providers, and MCP servers connect over stdio, HTTP or SSE.

Status

StatusActive
Views0
Outbound clicks0
Added10/21/2026

Platforms

Web AppMacOSWindowsLinux

Pricing

FreeFreemiumSubscription