Skip to content

Your AI team. At home on your Mac.

Recursive Intelligence brings coding, film and research into one workspace. Build with your choice of model, then review work that has been tested.

Get early accessSee pricing

Explore the app · Example of a completed build

Build a habit tracker with streaks and reminders

Built the habit tracker with streaks and reminders. All 31 tests passed; the changes are ready to review.

  • Read 3 symbols · src/App.tsx, useStreak, store.ts
  • Planned 4 steps
  • Built in a sandbox
  • Ran 31 tests · all passed
  • Checked it in a real browser
12 files changed +846 −0

Staged demo of the run view

Ready · habit-tracker
Add streaks and reminders
main3 files+84−12

Streaks count from the first check-in each day. Reminders use local notifications, so nothing leaves the Mac.

Sandbox run31 / 31
store.test.ts — streak resets after a missed day
date.test.ts — dayKey across time zones
App.test.tsx — reminder toggle persists
AnthropicOpenAIGooglexAIMoonshotAlibabaZ.aiDeepSeek

The teammate that makes everything

Build run · 3/3 steps
0:04Parse project · 412 symbols
0:06Pick context · 3 symbols
0:21Write change · 3 files
0:34Run tests in sandbox · 31 passed
0:35Diff ready for review
Open in Library

Tests its own work in a sandbox

Every change runs against your tests before you see the diff. You review finished work, not attempts.

Search models
Claude Opus 5.5
Claude Fable 5.1
GPT-6 Astra
Gemini 3.8 Flash
Grok 4.6
Kimi K3
DeepSeek V4.1 Flash

Gets the best from every frontier model

Claude for one turn, GPT for the next. Same thread, same files, same context.

Films with twenty models

Ten image and ten video models in one Studio panel. Every take lands in your Library.

The Research page in the app

Researches with a paper trail

Primary sources, cross-checked, in a report where every claim points to where it came from.

Original games. Ready to play.

Bramble Rally

Bramble Rally

A woodland kart racer.

Play game Sable Summit

Sable Summit

A moonlit alpine descent.

Play game Luma Grove

Luma Grove

A forest drop-and-merge puzzle.

Play game

From idea to finished work,
Recursive Intelligence does the whole job.

01

Understand the project

A Rust engine parses the code first and builds a symbol graph, so the model gets the three definitions a task needs instead of the whole repo.

Explore Build
Context manifest
412 symbols indexed · 3 sent src/store.ts useStreak() whole src/App.tsx <Habits /> signature src/date.ts dayKey() whole prompt prefix cached
02

Write and test

Every change runs against your tests in a sandbox before you see it. Failures go back to the agent, not to you.

See the benchmarks
Y
You

Add streaks and reminders to the habit tracker, and test it.

RI
Recursive Intelligence

Done. Three files changed, 31 tests passed in the sandbox. The diff is ready.

Worked for 35 s · +84 −12
03

Scan the surface

Point it at a site you own. It runs recon, port and web scans and hands you one report with what to fix first.

Explore Build
Scan · your-site.dev
Recon12 subdomains found
Port scan3 open ports · 443, 80, 22
Web scan1 header missing: CSP
Report ready · 1 item to fix
04

Automate the routine

Schedule a run and let it work while you don't — a morning briefing, a weekly review, a check when a page changes.

Explore Automations
“Every morning, brief me on what changed in my topic.”
Trigger
Every day 8:00
Action
Research the latest developments. Write a short briefing with primary sources, dates and open questions.
Model
Claude Opus 5.5

Connect once, use it everywhere
with the tools you already have

  • Open a pull request from a finished diff
  • Post the briefing to a channel
  • Turn an issue into a Build run
  • Build from a spec page
  • Read a frame before writing the UI
  • Research across your documents
  • Ship the preview when tests pass
  • Wire payments into a prototype

Measured,
not promised

Private by default
with controls you can see

  • Plugin tokens live in the macOS Keychain
  • Every Build change is tested in a sandbox first
  • Plugin traffic goes through one guarded client
  • Games run with no network access
  • Only the symbols a task needs are sent
  • You pick the model for every turn

Stop switching apps.
Do it all with Recursive Intelligence.

Get early accessSee pricing