Lucid AI
NewSay hello to Lycid Review

Where ideas come to life

LucidAI lets you build fully-functional apps in minutes with just your words.
No coding necessary.

Not sure where to start? Try one of these:
Core features

Everything you need to build faster

A chat-first agent that reads your repo, writes the code, runs the tests, and opens the PR — all streamed back to you live.

AI Software Engineer

Ship production code, autonomously.

Connect your GitHub or GitLab repo. Lucid AI reads the codebase, plans the change, writes the code, runs the tests, and opens a PR — all from a single chat turn.

Connect GitHub & GitLab repos
Understands your full codebase
Writes, tests & debugs code
Pushes commits & opens PRs
lucid-ai — feature/auth-flow
1
async function deployFeature(config) {
2
const branch = await git.createBranch();
3
const changes = await ai.implement(config);
4
// Running 24 test suites...
5
await git.commit('feat: add auth flow');
6
await git.pushAndCreatePR();
7
}
PR #247 created → feature/auth-flow → main
All 24 tests passing · 0 lint errors
How it works

Three ways to get there.

Start from scratch with zero setup, generate-then-export to your own Git, or hand Lucid the keys to a repo you already own.

Path 01

Zero-setup

No GitHub, no GitLab — Lucid AI provisions and configures everything for you.

  1. 1

    Describe your project

    Type what you want in chat. A landing page, a CRM, an internal tool — anything.

  2. 2

    Lucid spins up the sandbox

    A Docker workspace is created, dependencies installed, scaffolding generated.

  3. 3

    Iterate live in chat

    Add features, change copy, fix bugs. Preview updates instantly.

  4. 4

    Deploy to a Lucid domain

    soon

    One-click hosting on a managed domain.

Path 02

Generate then export

Build it with Lucid first, then move it into your own GitHub, GitLab, or Bitbucket.

  1. 1

    Describe your project

    Same chat-first flow — Lucid scaffolds and ships features in a sandbox.

  2. 2

    Connect your Git provider

    OAuth into GitHub, GitLab, or Bitbucket. Tokens are AES-256-CBC encrypted.

  3. 3

    Push to your repo

    Lucid creates the repo and pushes the initial commit. Branch + PR for changes after.

  4. 4

    Or download as ZIP

    soon

    Grab the project as a self-contained archive — no Git needed.

Path 03

Edit an existing repo

Bring code you already own. Connect a repo and let Lucid ship changes against it.

  1. 1

    Connect your Git provider

    Sign in with GitHub, GitLab, or Bitbucket. We only see the repos you pick.

  2. 2

    Pick the project

    Choose any repo you can write to. The agent clones it into its sandbox.

  3. 3

    Describe the change

    Add a feature, fix a bug, refactor a module. Tests run before the PR opens.

  4. 4

    Branch + PR ready for review

    You stay the reviewer. Merge when you’re happy.

Trust & isolation

Your code, your tokens, your boundary.

The agent runs in a sandbox. Your tokens are encrypted. Every row in our database is scoped to you. The architecture is the security story.

  • Tokens encrypted at rest

    · AES-256-CBC

    GitHub & GitLab PATs are AES-256-CBC encrypted before they hit the database. Only your session can decrypt them.

  • Per-session Docker sandbox

    · Resource-capped

    Every chat gets its own container with resource limits and an isolated workspace. Containers are destroyed when the session ends.

  • OAuth via Supabase

    · Google · GitHub · GitLab

    Sign in with Google, GitHub, or GitLab. We never see your password; the JWT is what authenticates every API and DB call.

  • Row-level data isolation

    · Postgres RLS

    Postgres RLS scopes every row to its owner via auth.uid(). There is no shared multi-tenant table to leak across.

Your session
Supabase JWT · auth.uid()
Encrypted at rest
Git PAT · AES-256-CBC
a3f4…b8c0 · b1de…7f23 · 9e84…a201
Per-session container
Docker sandbox · bind-mount
agent · running
boundary
The stack

Built on tools you already trust

Repos via GitHub or GitLab. Sandboxed execution in Docker. Auth and data on Supabase. Model routing across Anthropic, OpenAI, and Google.

01 · Code

Code & Sandbox

OAuth in, branch + PR out. The agent runs in an isolated container.

GitHub
OAuth · repos · PRs
GitLab
OAuth · repos · MRs
Docker
Per-session sandbox
02 · Auth

Auth & Database

OAuth-based identity, row-level isolation via Postgres RLS.

Supabase
Auth · Postgres · RLS
03 · LLM

LLM routing

LiteLLM under the hood — pick the model that fits the task.

Anthropic
Claude · default
OpenAI
GPT · o-series
Google AI
Gemini
04 · Auto-detected

Auto-detected stack

No config — the agent reads your manifest and picks the right tools.

Next.js
Sites · marketing
React
Dashboards · admin
Vue.js
Dashboards · admin
FastAPI
APIs · backends
Templates

Start from a template

Pick a starting point. Lucid customizes it to your idea — content, styling, and integrations included.

Use cases

What people actually do with it

From shipping a feature in one chat turn to clearing a stale backlog — Lucid hands you a branch and a PR, not a code editor to babysit.

Ship features from a prompt

  • Describe the change in plain English
  • Agent reads your codebase, plans, edits
  • Tests run before the PR opens
  • Branch and PR appear in your repo
lucid-ai · acme/marketing-site
live
Add Stripe checkout to /pricing.
Plan · 4 steps
▸ read app/pricing/page.tsx
▸ edit app/pricing/page.tsx
▸ add app/api/checkout/route.ts
▸ shell npm test --silent
✓ 18 tests passing
PR #312 · feat/stripe-checkout

Repo-aware agent

  • Reads your codebase before acting
  • Honors existing conventions & lint rules
  • Tool-call diffs you can review in chat

Sandboxed execution

  • Per-session Docker workspace
  • Resource limits + orphan cleanup
  • Code runs nowhere near your prod

Streaming events

  • WebSocket from the agent runtime
  • Plan, tool calls, test output — all live
  • Stop or redirect the agent mid-task

Ready when you are.

Drop a task here. We'll spin up a sandbox, write the code, run the tests, and open a PR on your repo.

Enter to send