Use Your Own API Keys on NeonCodex—Pay Nothing Extra
Most AI platforms lock you into their pricing. NeonCodex's BYOK feature lets you bring Claude, GPT-5, or Gemini and pay only what you use—directly to OpenAI or Anthropic. Here's exactly how to set it up.
The Problem: Middleman Pricing That Wastes Money
You've probably noticed: every AI platform marks up API costs. You pay OpenAI $0.003 per 1K input tokens for GPT-4, but use it through a third-party tool and suddenly you're paying 2-3x more. NeonCodex solved this with BYOK—Bring Your Own Key—which means you connect your OpenAI or Anthropic account directly and pay their prices, not a markup.
If you're already spending $50/month on OpenAI API calls, BYOK could cut that down to exactly $50 with no platform tax.
How BYOK Works (It's Simpler Than You Think)
Instead of NeonCodex hosting the model, you authenticate your own OpenAI or Anthropic account. When you run a prompt, it hits their API directly, charges your account, and NeonCodex handles the interface and features like Memory, Workflows, and Canvas on top.
You get two wins: (1) transparent, unmarkup pricing, and (2) access to NeonCodex's tooling—Memory, Knowledge Base, Prompt Library, Workflows—without paying the Pro plan fee if you only need BYOK.
Step-by-Step: Set Up BYOK in 3 Minutes
1. Get your API key
Go to platform.openai.com/api-keys (for OpenAI) or console.anthropic.com (for Anthropic). Click "Create new secret key." Copy it—you'll only see it once.
2. Add it to NeonCodex
Log into neoncodex.io, open Settings → API Keys. Paste your key in the OpenAI or Anthropic field. Name it something useful like "My OpenAI GPT-5" so you remember what it's for.
3. Start using it
When you open a new chat or workflow, select your BYOK key from the model dropdown instead of NeonCodex's hosted version. That's it.
Real Numbers: When BYOK Saves You Money
Let's say you're a solo developer building a chatbot with 200 API calls per week.
Option A: NeonCodex Pro ($2,499/month) — Unlimited calls, all models, but ₹2,499 whether you use it or not.
Option B: BYOK + Free NeonCodex tier — 10 tasks/day free with Qwen3 Coder, then use your own GPT-5.5 key for the rest. GPT-5.5 costs roughly $0.10 per 1K tokens (OpenAI pricing). Your 200 weekly calls with ~500 tokens each = 100K tokens/week = ~$10/week = ~$40/month. Total: $40, zero markup.
For heavy users (2,000+ calls/week), BYOK is the only financially sensible option.
Which Models Can You Bring?
OpenAI: GPT-4, GPT-4 Turbo, GPT-5.5 (with Pro account), o1 (with Pro)
Anthropic: Claude 3.5 Sonnet, Claude 3 Opus, Claude 3 Haiku
NeonCodex also hosts its own versions—Claude Sonnet 4.6, Claude Opus 4.8, Gemini 3.1 Flash, DeepSeek V4, MiniMax M3—so you have options depending on cost vs. performance tradeoffs.
Pick BYOK for models you already have budget for. Use NeonCodex's free Qwen3 Coder tier for quick experiments.
Three Smart Ways to Use BYOK
Use case 1: Batch processing
NeonCodex's Batch Processing feature works with BYOK keys. Process 1,000 prompts overnight using your OpenAI key, pay OpenAI's batch discount pricing (50% cheaper), and let NeonCodex handle the queue management. You save both on per-call costs and on platform fees.
Use case 2: Workflows with mixed models
Build a workflow that starts with Claude 3.5 Sonnet (your BYOK key), sends output to GPT-5.5 (your other BYOK key), then synthesizes with DeepSeek V4 (NeonCodex's hosted model). Each step charges exactly what it should—no platform markup anywhere.
Use case 3: Team access without paying per-seat
Share a NeonCodex workspace with teammates, but have one person manage the API keys (preferably your billing owner). Everyone uses the same Anthropic key, so you're not spinning up separate accounts. Some teams call this "key pooling"—it works, though you lose per-person usage tracking.
One Gotcha: Key Security
Don't commit your API key to GitHub or paste it in Slack. If you're sharing a workspace, set your OpenAI or Anthropic key to have spending limits (both platforms support this). On OpenAI, go to Billing → Usage limits and set a monthly hard cap. On Anthropic, use the "Set budget" feature in the console.
If someone compromises your key, revoke it immediately and generate a new one—takes 10 seconds.
What About the Free Tier?
NeonCodex's free plan gives you 10 tasks/day with Qwen3 Coder and Gemma 4. That's genuinely useful for testing workflows or prompts. If you burn through 10 tasks and need more, BYOK picks up the slack—you're not forced into Pro.
Many teams run: free tier for light work, BYOK for production, and skip Pro entirely.
The Real Win
BYOK is honest pricing. You pay OpenAI's rates, not inflated platform markups. Combined with NeonCodex's Memory feature (which learns your preferences and contexts across sessions) and Workflows, you get power-user tools without the power-user price tag.
Right now: Go grab your OpenAI or Anthropic API key from your console, head to neoncodex.io, and add it to Settings → API Keys. Then run one prompt with your own key and watch the charge hit your OpenAI/Anthropic account (not a mystery platform). That clarity is the whole point.