Every feature,
honestly explained
Every capability listed here is either live today or clearly marked. Exact limits, no vague promises.
Every frontier model in one place
Free plan includes Qwen3 Coder 480B, Gemma 4 31B, and Nemotron Super 120B. Pro unlocks Claude Sonnet/Opus 4.6, GPT-4o, GPT-5.5, Gemini 3.1 Pro (2M context), and more. Smart auto-routing picks the best model per task type automatically.
- ✓10 free tasks/day on free plan
- ✓50+ tasks/month on Pro
- ✓Automatic model selection
- ✓Extended thinking mode on Claude + GPT
AI that remembers your work
NeonCodex stores a summary of every completed task as a memory. Before each new task, relevant past memories are recalled and injected as context — so the AI knows your coding style, past decisions, and project context without you repeating yourself.
- ✓Automatic memory creation
- ✓Semantic similarity search
- ✓200 memories per user
- ✓Visible in Settings → Data
Upload and analyze any file
Upload images (with vision analysis), PDFs, CSV, code files, markdown, JSON, and more. The AI reads the full content and includes it as context. Images are sent as base64 to vision-capable models. Text files are extracted and prepended to your prompt.
- ✓Images: vision models see them
- ✓PDFs: text extracted automatically
- ✓CSV: data analysis + charts
- ✓Up to 100 MB per file, 5 files per task
Fetch and analyze any web page
Paste a URL into any prompt and NeonCodex automatically fetches the page content using a headless Playwright browser. Full JavaScript rendering — works on SPAs, dashboards, and dynamic sites. Use the /browser page for manual web research with AI analysis.
- ✓Automatic URL detection in prompts
- ✓Full JS rendering via Playwright
- ✓Up to 2 URLs per task
- ✓Manual browser at /browser
Execute Python in a secure sandbox
The AI can write and run Python code in an isolated subprocess with a 15-second timeout. Standard library + math, statistics, pandas-compatible operations. Output (stdout/stderr) is captured and shown inline. Dangerous operations (os.system, file writes, network) are blocked.
- ✓Isolated subprocess execution
- ✓15s timeout per run
- ✓Standard library available
- ✓Blocked: os.system, socket, file writes
Connect your own data sources
Add URL connectors (fetched via Playwright on refresh) or custom text/note connectors. Toggle "Auto-include" to prepend connector content as context for every task automatically. Perfect for adding company documentation, personal notes, or live data to every AI request.
- ✓URL connectors (live refresh)
- ✓Custom text/notes
- ✓Auto-include toggle
- ✓Settings → Data Sources
Bring your own API keys
Add your own OpenAI, Anthropic, OpenRouter, or Google API keys in Settings → API Keys. When set, your key is used for all your tasks instead of the platform key — using your own quota and billing. Keys are stored encrypted. Permanent nck_ API keys for programmatic access are in beta.
- ✓OpenAI, Anthropic, OpenRouter, Google
- ✓Stored encrypted in DB
- ✓Your own billing/quota
- ✓nck_ permanent API keys — Beta
Multi-turn AI conversations
Create persistent conversations that maintain full context across messages. Previous messages are sent to the model on every turn. Switch models mid-conversation. All conversations stored in your account history.
- ✓Full message history per turn
- ✓Switch models mid-conversation
- ✓Persistent across sessions
- ✓Dashboard → Conversations
Continue and extend any task
Every completed task has a "Continue" input below the output. Follow-ups are linked via parent_task_id and displayed as a full threaded conversation — exactly like Claude and ChatGPT. The full output of the original task is passed as context for the follow-up.
- ✓Thread linked by parent_task_id
- ✓Full output passed as context
- ✓Cryptographically random task URLs
- ✓Shareable via public share link
Share outputs publicly
Click the share button on any completed task to get a public URL. Anyone can view the output without logging in. Share links are cryptographically random (24-byte base64url). Disable with one click. Shared pages include a CTA to sign up.
- ✓Public /share/[token] URLs
- ✓No login required to view
- ✓Disable instantly
- ✓Views tracked per share
Speak your prompts
Click the microphone button to dictate your prompt. Uses Web Speech API on Chrome/Edge/Safari for real-time transcription. Falls back to MediaRecorder + OpenAI Whisper on Firefox and other browsers for maximum compatibility.
- ✓Web Speech API (Chrome/Edge/Safari)
- ✓Whisper fallback (Firefox)
- ✓Live interim transcription
- ✓Works on mobile browsers
Generate images with DALL-E 3
Two image capabilities: (1) Image Generation — use /imagine or "draw…" to create images with DALL-E 3. (2) Image-to-Code — upload any design screenshot, wireframe, or Figma export and Claude Vision generates complete HTML/CSS/React code that matches it pixel-accurately.
- ✓DALL-E 3 image generation (/imagine)
- ✓Image → HTML/CSS/React scaffolding
- ✓Claude Sonnet 4.6 Vision model
- ✓Live preview in Canvas panel
Collaborate with your team
Create team workspaces with roles (owner, admin, member, viewer). Invite members by email with 7-day token links. Organize work into shared projects. Private knowledge bases searchable by AI. Full audit log of all team actions.
- ✓Roles: owner, admin, member, viewer
- ✓Email invite links (7-day expiry)
- ✓Shared projects + tasks
- ✓Private knowledge base per team
Export your entire history
Download all your tasks and outputs as a Markdown file. Includes prompts, full outputs, model used, timestamps, and error messages. Settings → Data & Export → Download tasks.md.
- ✓Up to 500 tasks exported
- ✓Markdown format
- ✓Full output included
- ✓Settings → Data & Export
Ready to build?
Start free with 10 tasks/day. No credit card needed. Upgrade to Pro when you're ready.