Why plan first
- Long-horizon work needs a map. An agent that improvises for hours drifts. pre.dev’s spec — stack, milestones, user stories, acceptance criteria — is the contract every sprint builds and verifies against.
- Memory across sessions. The plan, the codebase, and everything the agent has learned about your project persist. Fork a session to try something, merge back what works.
- Autonomy you can steer. Run the whole roadmap on autopilot, review one task at a time, or dial effort per sprint — from a fast direct pass to the full research → code → verify pipeline.
What’s on this site
- Coding Agent Flagship The full loop — plan, build, verify, ship. Use it from the web workspace or the CLI.
- Browser Agents — fast, cheap, sandboxed browser automation for humans and AI. One REST call or MCP tool: scrape authed dashboards, fill forms, run multi-step flows, extract structured JSON.
- Architect API — the planning brain as a standalone REST API and MCP server. Drop
fast_spec/deep_specinto Cursor, Claude Code, or your own tooling. - CLI — the same agent, workspace, and slash commands in your terminal.
Find your way around
Coding Agent
Ship entire applications from a prompt. Real PRs, verified acceptance criteria, live preview and hosting, autonomous multi-hour builds.
Quickstart
Sign up, describe an app, hit Build on Autopilot. Minutes to a deployed product.
CLI
The web workspace’s chat, slash commands, effort and model controls — in your terminal, on your machine.
Browser Agents
Browser automation that behaves like an API. Sync, async, or SSE-streamed. Structured data out, not raw HTML.
Architect API
One
POST /fast-spec or /deep-spec returns a full structured plan — tech stack, milestones, user stories, granular subtasks. REST + MCP.Plans & Credits
What credits pay for, what each plan includes, and what happens when you run out mid-build.

