Describe your business, get a site or app
The OmniSpec OS v17 kernel runs all 9 stages and returns the packaged project with its main files and a preview link.
Runs on your machineRuns on your computer · no cloud required
Say what you want. NEURA builds it — and tells you the truth at every step.
One plain-language brief becomes a working site or app, a cinematic vertical video, ad creative and a markdown notes vault. Anything that leaves your machine only leaves with your own keys and your confirmation.
9pipeline stages
200prompts included
115languages detected
7social connectors
What it is
NEURA is a local-first business autopilot: it turns a plain-language brief into a usable site or app project, a cinematic vertical video, ad packs and a markdown notes vault — and reports every step as done, blocked, skipped or failed.
What this application is for
Eight concrete things you can do today, each tied to a real capability and to what it needs in order to run.
Get a vertical video from the same brief
A 9:16 cut for reels and shorts, rendered locally from the same request — no editing timeline involved.
Runs on your machineKeep everything you know in one vault
Markdown notes with [[wikilinks]], backlinks, a graph, tags and a daily note. The Obsidian that also does your chores.
Works offlineWrite ads that fit every platform
7 platforms (Meta, Google Search, PMax, TikTok, YouTube, LinkedIn, X) across 6 angles, with the real character limits enforced locally.
Works offlinePut the site or app online
Deploys to Netlify or the portfolio, then verifies the live URL over HTTP — if it does not open, it says it failed.
Needs a keyPublish what you produced to your networks
Real connectors behind a credential gate: without the platform token the job is reported blocked with the exact variable name and zero network calls.
Needs a keyReach your leads, in their language
Personalised first contact with recorded consent per lead and a STOP footer. Without SendGrid the message is blocked — never "sent".
Needs a keyWake up to what actually happened
A daily report: what went up, what failed and what needs your decision. No invented reach or return numbers.
Works offlineSeven pillars, one button
The autopilot runs all seven in one go. Each pillar is isolated: one failure never hides the others, and each one says what it did and what was missing.
vault
Notes vault
Writes the brief as an idea note and a run note with the outcome.
Needs: nothing
apps
Apps and sites
The kernel creates the project and returns a preview to share.
Needs: kernel on 3847
video
Video
Renders a cinematic vertical cut from the same brief.
Needs: kernel on 3847
ads
Ads
A creative pack per platform and angle, with limits enforced locally.
Needs: LLM key (improves copy)
site
Site publish
Deploys and confirms the live URL over HTTP.
Needs: NETLIFY_AUTH_TOKEN
social
Social
A real attempt per platform, with a credential gate and an idempotency receipt.
Needs: each platform's token
email
Personalised first contact, in the lead's language, with consent.
Needs: SENDGRID_API_KEY + NEURA_EMAIL_FROM
| Service | Variable | Where to get it | State |
|---|---|---|---|
| SendGrid | SENDGRID_API_KEY |
app.sendgrid.com → Settings → API Keys | blocked until pasted |
| Netlify | NETLIFY_AUTH_TOKEN |
app.netlify.com → User settings → Applications | blocked until pasted |
| DeepSeek | DEEPSEEK_API_KEY |
platform.deepseek.com → API Keys | optional |
| Google Maps | GOOGLE_MAPS_API_KEY |
console.cloud.google.com → Places API | optional |
| TikTok | TIKTOK_ACCESS_TOKEN |
developers.tiktok.com → Login Kit | blocked until pasted |
| Twilio | TWILIO_ACCOUNT_SID |
console.twilio.com → Account Info | optional |
| Google AI Pro + Gemini API | GEMINI_API_KEY |
aistudio.google.com/apikey | optional |
| Google Cloud + billing | GOOGLE_CLOUD_PROJECT |
console.cloud.google.com/projectcreate | needed for paid APIs |
| Google Veo (video) | GOOGLE_CLOUD_PROJECT + GOOGLE_VEO_MODEL |
ai.google.dev/gemini-api/docs/video | billed per second |
| Google Imagen (images) | GOOGLE_IMAGEN_MODEL |
ai.google.dev/gemini-api/docs/image-generation | billed per image |
| Google Workspace (Drive, Docs, Sheets, Gmail) | GOOGLE_CLIENT_ID + GOOGLE_CLIENT_SECRET + GOOGLE_REFRESH_TOKEN |
console.cloud.google.com/apis/credentials | OAuth: consent screen + token |
| NotebookLM Enterprise | NOTEBOOKLM_SA_JSON + NOTEBOOKLM_PROJECT |
cloud.google.com/generative-ai-app-builder/docs/notebooklm | needs an Enterprise licence |
Every row has an assistant with 3 to 4 steps and a direct link to get the key. No key leaves your computer: they stay in the local .env file.
Optional
Google AI Pro (optional, for clients)
NEURA works without it. If you already have the Google AI Pro plan, connecting the Google AI Studio key gives you more quota and unlocks Google's paid tools inside the app.
More Gemini quota
Copy, translations and the notebook use Gemini with your account limits.
Video with Veo
A cinematic cut generated by Google, as an alternative to the local render.
Images with Imagen
Product shots and ad backgrounds from a description.
NotebookLM
NEURA prepares the sources; with Google, the notebook lives in your account.
- Open aistudio.google.com/apikey with the same Google account as the plan.
- Create the key and paste it in Integrations → Google AI Pro (inside the app).
- For Veo and Imagen, create a Google Cloud project with billing enabled.
One key is enough: GEMINI_API_KEY. The Google Cloud project is for the paid APIs (Veo, Imagen, NotebookLM Enterprise).
Without this key NEURA keeps working: text falls back to the local/deterministic mode and says it does not know instead of inventing. The plan is billed by Google, never by NEURA.
aistudio.google.com/apikeyNotebook: the knowledge that makes agents run faster
Ask over what NEURA knows and get an answer grounded in your notes, with the note cited. That same knowledge is then handed to OmniSpec OS v17 and to NotebookLM.
Answers with citations
The answer comes only from your notes, with the note it used listed underneath. With no source it says it does not know instead of inventing.
A brief in milliseconds
A local summary of what the goal needs (no AI request spent) that agents receive before they start working.
Handed to OmniSpec OS v17
The brief goes into the kernel memory and the orchestrator splits the goal across agents — more work done, less context repeated.
A pack for NotebookLM
One markdown file with your notes as sources, plus instructions and suggested questions, ready to drop into notebooklm.google.com.
Honest note: the consumer NotebookLM has no public API. NEURA prepares the source pack (and NotebookLM Enterprise becomes available once you connect the Google Cloud credentials).
200 ready prompts, two directions
Not sure what to ask? The library has 200 prompts in 10 categories across two tracks — plus a generator that writes prompts for your company specifically.
Categories
Start a company from zero115 prompts
Grow the company you have99 prompts
- Site
- Brand
- Content
- Ads
- Video
- SEO
- Leads
- Operations
- Sales
And it is not static
With an AI key connected, NEURA writes new prompts for your company — name, sector, city and language — validates them, drops duplicates and stores them per company (up to 120 per company). Without an AI key the generation is blocked and names the missing variable; it never invents prompts.
Samples from the library
Create a restaurant website with menu, bookings and a map.
Write the brand identity: tone of voice, palette and three core messages.
Generate 6 Instagram ad variations within the real character limits.
Make a 20-second vertical video that presents the product in three scenes.
Write the title, description and keywords for the home page.
Write the first email to clients in a nearby city, with a STOP footer.
Find 20 businesses in my sector and city and prepare the outreach.
Create the shop service guide in 10 rules for the new team.
Write a one-page commercial proposal with pricing and terms.
Plan 30 posts for a month, with themes and captions.
Build a simple online store with a catalogue and checkout.
Improve the current logo while keeping its shape and modernising the line.
All 200 prompts are inside the ZIP, in markdown, ready to copy.
Local install in three steps
No database to install, no server of ours. Download the ZIP, run it with Node and open the browser.
-
Download and unzip
Save the ZIP in your own folder and extract everything into a new one.
neura-v0.1.0-local.zip -
Start the app
On Windows a double click is enough. Anywhere else, two commands.
INICIAR.bat | node backend/server.ts + npm run web -
Open the browser
Type your business, read the plan, and only then confirm the run.
http://localhost:5310
Honest states, always
Every task ends in one of four states, with the real reason. That is the difference between a toy and a tool you can trust.
Done
It happened and there is real evidence (a file, a verified URL, a returned id).
Blocked
A credential or an approval is missing. It names the exact variable and makes no network calls.
Skipped
It did not apply: for example, a lead without recorded consent is never emailed.
Failed
It tried and could not, with the reason. Nothing is hidden or disguised as success.
- We never report invented reach, impressions, clicks or return.
- Nothing is sent or published without your explicit confirmation.
- Ad platforms require an approved business account: NEURA exports the pack and a deep link instead of pretending to publish.
- Cloud sync is planned, not available.
Frequently asked questions
What does NEURA do?
NEURA is a local-first business autopilot. You describe what you want in plain language and it produces a working website or app, a cinematic vertical video, ad creative packs and a markdown notes vault. With your own credentials it also publishes to your social accounts and emails your leads. Every step is reported as done, blocked, skipped or failed.
Does NEURA work offline?
The interface, the notes vault, the 200-prompt library and the ad-pack generator work offline. Site and video generation use the OmniSpec OS v17 kernel, which runs locally on your machine at 127.0.0.1:3847. Anything that leaves your machine needs internet and your own credential.
Which languages does NEURA support?
English is the base language, and Portuguese and Greek are fully translated in the interface. NEURA also detects around 115 languages for your brief and can write your site, video, ads and emails for a different audience language than the one you speak. The prompt library and text generated live use an AI key; without that key the text comes out in English.
Does NEURA need paid API keys?
No for the core. The vault, prompts, ad packs and local pipeline work without any key. Keys are only needed for actions that talk to an external service: SendGrid for email, Netlify for deploy, each social platform's token for publishing, Google Places for lead discovery, and an LLM key such as DeepSeek or Gemini for the smartest copy and in-app translation. When a key is missing, NEURA reports blocked with the exact variable name instead of inventing a result.
Where is my data stored?
On your machine. Data lives in local SQLite and JSON files, and the vault is markdown inside the NEURA folder. Folders you point NEURA at are read in place and never uploaded. Upload is only used for files that come from outside your computer.
How much does NEURA cost?
The ZIP is free to download and run. You only pay the third-party providers whose keys you choose to connect, such as SendGrid, Netlify or a social platform, and they bill you directly.
Is publishing automatic?
No. Nothing is sent or published without your explicit confirmation. Email requires recorded consent per lead. Every social platform has a credential gate, so without the platform's token the job is reported blocked with zero network calls. Ad-platform publishing is not possible today because those platforms require an approved business account, so NEURA exports the ad pack and links you to the ad manager instead of pretending to publish.
How do I install NEURA?
Download the ZIP, unzip it and run it with Node.js 22.6 or newer: node backend/server.ts for the API and npm run web for the interface. There is no database to install because it uses local SQLite and JSON files. The OmniSpec OS v17 kernel only needs to be running if you want NEURA to generate the site or app and render the video.
Create your NEURA account
Leave your details and we keep your contact as a lead — in the same place the leads of the other applications land. We answer by email with the access and the next step.
Talk to the person who built it
Questions, a concrete project, or a key you cannot find: write directly. I answer personally.