Documentation

Quick start

1. Connect AI. Click Connect with OpenRouter in AI settings. Ntekuma builds on your own OpenRouter credits.
2. Describe what you want. Write it in plain language on the start page. Your first message creates the project.
3. Pick an AI mode. Best quality, Balanced, or Low cost. Ntekuma routes each task to the right model.
4. Run the build. Follow the agent's progress and the live preview.
5. Iterate and refine. Request changes to polish your project.

What you can generate

• Web pages: static HTML/CSS/JS sites, landing pages, portfolios.
• UI prototypes: component-driven layouts and design systems.
• Dashboards: data visualization and analytics interfaces.
• WordPress and Joomla themes: previewed in the workspace and downloadable as theme files.
• Documents: reports, guides, checklists, and summaries, including cited academic writing.
• Storyboards: script-to-storyboard in Cinema, with exports for video tools.
• Design to code: upload screenshots or mockups and get React, Vue, or HTML code.

Prompting tips

Good prompts

• “Create a landing page for a coffee shop with a hero image, menu section, location map, and contact form. Use warm brown colors and modern typography.”
• “Build a dashboard showing sales metrics with a sidebar navigation, 4 KPI cards at the top, a line chart for revenue trends, and a data table below.”

Avoid these

• “Make me a website” (too vague)
• “Build an app like Instagram” (too complex for one prompt)

More tips

• Be explicit about the target: web page, theme, document, or storyboard
• List required sections or screens in order of importance
• Specify fonts, colors, and layout constraints when relevant
• For edits, reference the exact element and what should stay the same
• Upload reference images to show the style you want

Iterative editing

Describe the change, not the whole thing. When requesting changes, say what to keep and what to modify. This refines the project instead of rebuilding it.

Example: “Keep the hero layout, change the CTA button to red, and replace the feature icons with illustrations.”

Uploads and references

You can upload images, PDFs, or text files to guide the agent. Be specific about how each upload should be used.

Supported formats

• Images: PNG, JPG, WebP, SVG
• Documents: PDF, TXT, MD
• Data: JSON, CSV

Usage examples

• “Use this photo as the hero background”
• “Follow this layout exactly”
• “Extract the color palette from this image”

Connect AI (OpenRouter)

Ntekuma builds on your own OpenRouter credits, so you control your AI spend and Ntekuma never charges you for tokens. Your subscription pays for the platform.

1. Open AI settings and click Connect with OpenRouter. A small OpenRouter window opens over Ntekuma.
2. Sign in to OpenRouter (or create a free account), click Authorize, then add pay-as-you-go credit. The window closes by itself when Ntekuma sees the credit. Your key is created and stored for you, encrypted, so there is nothing to copy.
3. Need more credit later? Use Top up in AI settings or Usage & billing. Already have a key? You can still paste it in AI settings.
4. Start building. OpenRouter bills your own account.

Your key is AES-256-GCM encrypted and only decrypted in memory to run your AI tasks or show your OpenRouter balance.

AI modes and model routing

Instead of picking a model by hand, choose an AI mode. Ntekuma routes each task (planning, UI, backend, edits) to the right Claude model on your key.

Today Ntekuma routes to Claude models through OpenRouter.

Platform features

Version history

Every build is saved as a version. Compare versions, view diffs, and roll back to a previous state.

Templates

Start from a template such as a SaaS starter, e-commerce store, analytics dashboard, CRM, portfolio, or AI chat interface.

GitHub sync

Connect a GitHub repository to push changes and pull updates.

Custom domains

Publish a project to your own domain, with DNS verification and automatic SSL certificates.

Agent mode

Describe a goal and the agent plans and carries out multi-step work: it reads files, writes code, runs commands, and iterates until done.

Downloads and publishing

After a successful build, you can:

• Download a ZIP of all project files
• Download individual files from the file browser
• Publish to a public link or a custom domain

Cinema (storyboards)

Cinema turns a screenplay into a storyboard. Define characters once and they stay consistent from shot to shot.

How it works

1. Write your script. Import or write a screenplay with character descriptions, scene headings, and dialogue.
2. Generate storyboards. Ntekuma creates a preview image for each shot. Review and regenerate frames until they are right.
3. Export. Download a PDF storyboard, a pitch deck, a CSV shot list, a Fountain script, or prompt packs formatted for Sora, Veo, and Runway.

Camera settings

Every shot carries camera metadata. Ntekuma picks settings to suit the shot, or you can set them in your script.

• Lenses: 14mm ultra-wide to 200mm telephoto, anamorphic, tilt-shift, macro, fisheye
• Lighting: Rembrandt, butterfly, split, chiaroscuro, silhouette, golden hour, neon-noir, and more
• Composition: rule of thirds, golden ratio, center frame, leading lines, symmetry, layered depth
• Camera and film: ARRI Alexa 65, RED V-Raptor, Sony Venice 2 bodies; Portra 400, CineStill 800T, Vision3 500T film stocks

Script camera directives

Add camera directives directly in your script. The parser detects them automatically:

[SHOT: Close-up with 85mm lens, Rembrandt lighting, rule of thirds]
[Camera: Slow dolly push-in, anamorphic, golden hour]
[SHOT: Wide establishing, 24mm, chiaroscuro, Portra 400]

Without directives, Ntekuma chooses settings based on shot type, mood, and style.

Academic source search

For research writing, Ntekuma searches academic databases including OpenAlex, Semantic Scholar, and CrossRef, and grounds your text in real, verifiable sources.

How grounded generation works

1. You prompt. Write what you need: an essay, research paper, or literature review.
2. Ntekuma searches. It queries several academic databases in parallel for real papers.
3. Full text retrieved. It reads abstracts and, where available, full text from open-access sources and the publisher keys you add.
4. Grounded generation. The text is written from the actual findings in the papers.
5. Citations. Inline (Author, Year) citations and a formatted references section.
6. Verification. Claims are checked against source data for accuracy.

Access to papers

While writing, the agent can read a paper's full text, not just its abstract, before citing it. Open-access copies from PubMed Central, Europe PMC and Unpaywall are always used. To reach more full text, add your own Elsevier or Springer Nature API key (and, for Elsevier, your institution's token) in Research sources. An NCBI key there raises the PubMed Central request limit.

Keys are checked with the publisher when you save them, encrypted with AES-256-GCM, and used only by the server to fetch the articles. The AI model never sees them. Cookie-based institutional logins (JSTOR, Wiley and similar) are not supported.

Troubleshooting

• Preview is blank. Check that the entrypoint file (index.html) exists.
• Output looks wrong. Narrow your prompt and specify a single target.
• Build fails. Review the error logs and confirm OpenRouter is connected in AI settings.
• Build drifts from the original. Use explicit “keep this, change that” edits.
• Slow generation. Use the Balanced or Low cost AI mode while iterating, and Best quality for the final pass.

Need more help?

Contact support at support@ntekuma.com.