Documentation Index Fetch the complete documentation index at: https://docs.pre.dev/llms.txt
Use this file to discover all available pages before exploring further.
Pick the path that matches how you work: a one-line curl, an MCP install into your editor, or an SDK call.
1. Get your API key
Grab a key Visit the playground to copy your API key and see remaining credits.
2. Call the Architect
curl
MCP (Cursor)
MCP (Claude Code)
TypeScript SDK
Python SDK
curl -X POST https://api.pre.dev/fast-spec \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{"input": "Build a task management app with team collaboration"}'
You’ll get back a structured spec with milestones, user stories, and architecture — in ~30–60 seconds. Click the button, restart Cursor. On first use a browser tab opens — sign in to pre.dev and pick which account or organization to authorize. Then ask: Use pre.dev to generate a spec for a task management app with team collaboration
Full setup for Cursor / Claude Code / VS Code / Windsurf: MCP Setup . claude mcp add --transport http predev https://api.pre.dev/mcp
Your browser will open to authorize — pick which account or organization to connect. Then ask Claude: Use pre.dev fast_spec to plan a task management app
import { Predev } from '@predev/sdk' ;
const client = new Predev ({ apiKey: process . env . PREDEV_API_KEY });
const spec = await client . fastSpec ({
input: 'Build a task management app with team collaboration' ,
});
console . log ( spec );
See Node SDK for full docs. from predev import Predev
client = Predev( api_key = os.environ[ "PREDEV_API_KEY" ])
spec = client.fast_spec( input = "Build a task management app with team collaboration" )
print (spec)
See Python SDK for full docs.
3. Hand the spec to your coding agent
Whatever you got back is structured markdown (or JSON) your coding agent can consume directly. Paste it into Cursor’s chat, drop it into Claude Code, or feed it to Lovable / Bolt. Your agent now has the architectural context to build the whole thing on the first pass.
For a richer breakdown, use /deep-spec instead — it adds granular subtasks and per-task acceptance criteria (~2–5 minutes).
What’s next?
Fast vs Deep Pick the right depth for your project.
Understanding Specs What’s actually in a pre.dev spec.
Full API reference All endpoints, parameters, and response schemas.
MCP Setup Install into every major AI editor.