Skip to main content

Docs

Every Dracon surface on one page. Pick a product below, or read how AI works here: bring your own key, no relay, no surprises.

Product guides

Where to go

Books A full shelf to browse. Open one and start reading. Music Library, playlists, and search across the catalog. Games Released games, playable in the browser. Chat Conversations powered by your own provider keys. Tools VPS, storage, email, AI sandboxes: real prices, no signup. AI Hub Rankings, providers, free lane: bring your own key. How it works below.
AI on Dracon

Bring your own key

Dracon's AI features are BYOK (bring your own key): you sign up directly with an AI provider, paste the provider key into AI Hub, and your browser talks to the provider directly. Your browser (not a Dracon server) holds the key in memory and calls the provider from your machine.

There is no Dracon-hosted relay endpoint to call: earlier API-key-based access (drn_ keys against /api/v1/ai/chat/completions) was removed on 2026-08-25. If you maintain a Chrome extension that used it, switch it to the user's own provider key stored locally (e.g. chrome.storage.sync) calling the provider's native endpoint.

Open AI Hub

AI providers (48)

Providers you can bring a key for in AI Hub. The full catalog lives in web/packages/byok/src/data/byok-registry.json, the source of truth; this list is its generated view at web/ai-hub/src/lib/data/ai-providers-catalog.json.

  • Z.AI GLM free tier Lite: 10,000 credits / week. Pro: 6× Lite usage. Max: 14× Lite usage. The plan supports 20+ agent tools; exact model access and limits can change.
  • Agnes AI Starter: 1,500 model requests / 5-hour window. Plus: 7,500 / 5h. Pro: 30,000 / 5h. Higher RPM on paid tiers; ~100 TPS typical, ~150 off-peak.
  • Mimo free tier One-time 50-credit trial for the Airo builder.
  • MiniMax M Plan usage is tiered: Go is the base allowance, Explore includes 3x Go usage, and Build includes 7.5x Go usage. The public pricing page publishes relative multipliers rather than absolute monthly limits.
  • OpenAI Codex free tier Usage depends on plan, model, surface, task complexity, context, reasoning, speed, and tools. Pro $100 provides 5x higher usage than Plus; Pro $200 provides 20x usage than Plus. Allowances and reset times are shown in Settings, the usage dashboard, or /status.
  • Alibaba Coding Plan free tier Coding Plan Pro: up to 6,000 requests / 5 hours, 45,000 / week, and 90,000 / month. New-user free quota is a 10M-token trial (see product page).
  • Cline free tier Cline provider: per-model limits (rotate). BYOK: provider-specific.
  • Qwen Cloud Token Plan Token Plan quota varies by tier and off-peak window
  • OpenRouter free tier 50 req/day on the base free tier; $10 in PAYG credits unlocks 1,000 req/day of free models (model pool rotates)
  • Google AI The reviewed UK storefront advertises 2x (Plus), 4x (Pro), 5x (Ultra 5x), and 20x (Ultra 20x) access; limits and availability vary by feature and country.
  • Ollama free tier Local: unlimited. Cloud PAYG $0: 3 concurrent requests.
  • Synthetic Varies by plan
  • LLM Gateway free tier Free models capped at 20 reqs/min; paid models not rate limited (pricing page, checked 2026-10-27).
  • StepFun Varies by plan
  • BytePlus ModelArk free tier Trial quota for new users; paid per-token rates after.
  • Grok Usage-based, no minimums; per-model rate limits apply (see xAI docs)
  • DeepSeek See platform.deepseek.com; paid tiers advertise quota per 5h window.
  • AI Router Free: 1K req/mo, Starter: 20K req/mo, Pro: 100K req/mo
  • Kilo Code free tier Per-model request limits rotate; see the Kilo pricing page and docs.
  • Venice Varies by plan
  • Anthropic Consumer usage resets on a rolling five-hour session window; paid plans add weekly limits. Capacity depends on conversation length, model, features, and task complexity, so there is no fixed message count.
  • Cerebras 1M tokens/day
  • Chutes Varies by model
  • Cloudflare Workers AI free tier 10K neurons/day free allocation; $0.011 per 1K neurons overage.
  • CrofAI 500 requests/day (Hobby)
  • DeepInfra 200 concurrent requests
  • Fireworks AI High rate limits with postpaid billing
  • GitHub Models Service retired; see Azure AI Foundry for new projects.
  • Groq free tier Free Plan Limits table (e.g. 30 RPM with up to 14.4K RPD); Developer plan higher; 429s carry retry-after.
  • Hugging Face No free tier: inference credit must be purchased.
  • Hyperbolic Pro plan required for inference (one-time $5+ deposit).
  • Mistral Le Chat free tier Le Chat (consumer): messages, web searches, libraries, and image generation scale by tier (Free = base; Pro = up to 6x base messages, 5x base web searches, 15GB libraries, 40x base image generation; Team = Pro per user + admin + domain verification). La Plateforme (API): pay-per-token, separate from these tiers.
  • Google AI Studio free tier Free Tier for new accounts; Paid Tier per-1M-token rates after (see pricing).
  • Nous Portal Ultra: 1,600 RPM
  • NVIDIA NIM free tier ~40 RPM
  • OVHcloud AI Endpoints free tier Keyless catalogue trial is rate-limited; paid keys raise limits.
  • Perplexity ~50 RPM (tiered by spend)
  • Replicate 6 req/min free, 3K RPM with payment
  • RunPod Varies by plan
  • Novita AI
  • SambaNova Generous dev tier
  • Scaleway 1M free tokens (permanent)
  • SiliconFlow free tier Free chat models about 100 RPM with fixed TPM caps; paid models scale by monthly spend tier.
  • OpenCode free tier Go publishes estimated requests per 5 hours and monthly usage by model; limits rotate and vary by model. Zen is pay-as-you-go with configurable workspace monthly limits and auto-reload.
  • Muse Code Everyday Usage: 10-50 prompts every 5 hours. High Usage: 5x Everyday Usage. Power Usage: 20x Everyday Usage. Limits refresh; Power plan holders wait for refresh at the limit.
  • OpenDesign Go: 10 concurrent tasks. Plus: 20. Pro: 30. Max: 30. One monthly credit ceiling per account; the app and the API draw on it together.
  • Kimi Varies by usage tier
  • OpenAI API