Claude Code hits rate limits. Accounts get locked. Phone verification fails.
You're paying $200/month and still can't work.
brcc removes all of that. Any model. No limits. Pay what you use.
brcc is a local proxy that lets you run Claude Code with any LLM model — GPT-5, Claude, Gemini, DeepSeek, Grok, and 50+ more — without rate limits, without an Anthropic account, and without phone verification. You pay per request with USDC via the x402 protocol. Your wallet is your identity. Your private key never leaves your machine.
Claude Code users are frustrated. Over 4,350 GitHub issue comments about the same problems:
Every one of these people is a potential brcc user.
brcc eliminates all of these problems:
- No rate limits — pay per request, use as much as you want
- No account — a wallet is generated locally, no signup
- No phone verification — USDC is your authentication
- No region restrictions — works everywhere, priced the same
- No billing surprises — transparent per-token pricing
| Claude Max ($200/mo) | OpenRouter | brcc | |
|---|---|---|---|
| Rate limits | Constantly hit | Per-model limits | None |
| Account required | Yes + phone | Yes + email | No |
| Models | Claude only | 200+ (manual select) | 50+ (auto or manual) |
| Payment | Credit card, subscription | Credit card, pre-pay | USDC per-request |
| Auth | OAuth + API key conflicts | API key | Wallet signature |
| Pricing | Opaque | Transparent | Transparent |
| Runs locally | N/A | No | Yes (proxy) |
| Monthly cost | $100-200 fixed | Varies | $5-50 based on usage |
curl -fsSL https://blockrun.ai/brcc-install | bashInstalls Node.js (if needed) + Claude Code + brcc + creates wallet.
# 1. Install Claude Code
curl -fsSL https://claude.ai/install.sh | bash
# 2. Install brcc (use sudo on Linux)
sudo npm install -g @blockrun/cc # use sudo on Linux
# 3. Create wallet
brcc setup base # Base chain (Coinbase L2)
# or
brcc setup solana # Solana chain
# 4. Fund your wallet with USDC (or use free models without funding)
# 5. Launch
brcc startThat's it. Claude Code opens with access to 50+ models, no rate limits.
brcc start # Default: smart routing (blockrun/auto)
brcc start --model blockrun/eco # Cheapest capable model
brcc start --model blockrun/premium # Best quality
brcc start --model blockrun/free # Free tier only
brcc start --model deepseek/deepseek-chat # Direct model access
brcc start --model anthropic/claude-opus-4.6 # Most capablebrcc includes ClawRouter's 15-dimension classifier for automatic model selection:
| Profile | Strategy | Savings | Best For |
|---|---|---|---|
blockrun/auto |
Balanced (default) | 74-100% | General use |
blockrun/eco |
Cheapest possible | 95-100% | Maximum savings |
blockrun/premium |
Best quality | 0% | Mission-critical |
blockrun/free |
Free tier only | 100% | Zero cost |
How it works:
"What is 2+2?" → SIMPLE → gemini-flash ($0.0002)
"Write a React component" → MEDIUM → kimi-k2.5 ($0.002)
"Design a microservice..." → COMPLEX → gemini-3.1-pro ($0.007)
"Prove this theorem..." → REASONING → grok-4-fast ($0.0004)
In-session switching:
use auto # Switch to smart routing
use eco # Switch to cheapest
use premium # Switch to best quality
use free # Switch to free models
use sonnet # Direct Claude Sonnet
use gpt # GPT-5.4
use codex # GPT-5.3 Codex
use o3 # OpenAI o3
use gemini # Gemini 2.5 Pro
use flash # Gemini 2.5 Flash
use grok # Grok 3
use grok-4 # Grok 4
use deepseek # DeepSeek Chat
use r1 # DeepSeek Reasoner
use minimax # Minimax M2.7
use kimi # Kimi K2.5
use free # Nemotron Ultra 253B (free)
use nemotron # Nemotron Ultra 253B (free)
use devstral # Devstral 2 123B (free)
use qwen-coder # Qwen3 Coder 480B (free)
Use /model to switch between Sonnet, Opus, and Haiku. Each maps to the BlockRun model you've configured:
# Customize what each /model option routes to
brcc config set sonnet-model anthropic/claude-sonnet-4.6 # default
brcc config set opus-model anthropic/claude-opus-4.6 # default
brcc config set haiku-model deepseek/deepseek-chat # cheap alternative$ brcc models
Free Models (no USDC needed)
──────────────────────────────────────────────────────────────────────
nvidia/nemotron-ultra-253b
nvidia/gpt-oss-120b
nvidia/deepseek-v3.2
nvidia/mistral-large-3-675b
nvidia/qwen3-coder-480b
nvidia/devstral-2-123b
nvidia/llama-4-maverick
nvidia/glm-4.7
... (11 free models)
Paid Models
──────────────────────────────────────────────────────────────────────
Model Input Output
openai/gpt-5-nano $0.05/M $0.40/M
xai/grok-4-1-fast-reasoning $0.20/M $0.50/M
deepseek/deepseek-chat $0.28/M $0.42/M
minimax/minimax-m2.7 $0.30/M $1.20/M
google/gemini-2.5-flash $0.30/M $2.50/M
moonshot/kimi-k2.5 $0.60/M $3.00/M
anthropic/claude-haiku-4.5 $1.00/M $5.00/M
openai/gpt-5.3-codex $1.75/M $14.00/M
google/gemini-2.5-pro $1.25/M $10.00/M
google/gemini-3.1-pro $2.00/M $12.00/M
openai/gpt-5.4 $2.50/M $15.00/M
anthropic/claude-sonnet-4.6 $3.00/M $15.00/M
anthropic/claude-opus-4.6 $5.00/M $25.00/M
openai/gpt-5.4-pro $30.00/M $180.00/M
... (50+ models total)| Model | ~Requests per $5 | Best For |
|---|---|---|
| Nemotron Ultra 253B | Unlimited | Free tier |
| Grok 4.1 Fast | ~3,500 | Fast reasoning (budget) |
| DeepSeek V3 | ~5,000 | Budget coding |
| Gemini 2.5 Flash | ~1,200 | Balanced speed/cost |
| Kimi K2.5 | ~800 | Mid-range coding |
| Claude Haiku 4.5 | ~500 | Fast tasks |
| GPT-5.3 Codex | ~180 | Code generation |
| Claude Sonnet 4.6 | ~100 | General coding |
| GPT-5.4 | ~80 | Reasoning |
| Claude Opus 4.6 | ~50 | Most capable |
| Command | Description |
|---|---|
brcc setup [base|solana] |
Create wallet for payments |
brcc start [--model <id>] |
Start proxy + launch Claude Code |
brcc models |
List all models with pricing |
brcc balance |
Check wallet USDC balance |
brcc stats |
View usage statistics and savings |
brcc config set <key> <value> |
Configure model mappings |
brcc config list |
View current settings |
brcc setup # Default: Base chain
brcc setup base # Coinbase L2 — low fees, fast
brcc setup solana # Solana — also low fees, fastYour wallet is saved to ~/.blockrun/ and shared with all BlockRun tools.
brcc start # Default model
brcc start --model nvidia/nemotron-ultra-253b # Free model
brcc start --model openai/gpt-5.4 # Specific model
brcc start --no-launch # Proxy only mode
brcc start --no-fallback # Disable auto-fallback
brcc start -p 9000 # Custom portView your usage statistics and cost savings:
$ brcc stats
📊 brcc Usage Statistics
───────────────────────────────────────────────────────────
Overview (7 days)
Requests: 1,234
Total Cost: $4.5672
Avg per Request: $0.003701
Input Tokens: 2,456,000
Output Tokens: 892,000
Fallbacks: 23 (1.9%)
By Model
anthropic/claude-sonnet-4.6
450 req · $2.1340 (46.7%) · 245ms avg
deepseek/deepseek-chat
620 req · $0.8901 (19.5%) · 180ms avg
↳ 12 fallback recoveries
nvidia/nemotron-ultra-253b
164 req · $0.0000 (0%) · 320ms avg
💰 Savings vs Claude Opus
Opus equivalent: $34.62
Your actual cost: $4.57
Saved: $30.05 (86.8%)
───────────────────────────────────────────────────────────
Run `brcc stats --clear` to reset statistics
$ brcc stats --clear # Reset all statistics
$ brcc stats --json # Output as JSON (for scripts)brcc config set default-model nvidia/nemotron-ultra-253b
brcc config set sonnet-model openai/gpt-5.4
brcc config set opus-model anthropic/claude-opus-4.6
brcc config set haiku-model deepseek/deepseek-chat
brcc config listWhen a model returns an error (429 rate limit, 500+ server error), brcc automatically retries with backup models. This ensures your work never stops.
Default fallback chain:
your-selected-model
↓ (if 429/500/502/503/504)
blockrun/auto (smart routing)
↓
blockrun/eco (cheapest capable)
↓
deepseek/deepseek-chat
↓
nvidia/nemotron-ultra-253b (free, always available)
How it looks:
[brcc] ⚠️ anthropic/claude-sonnet-4.6 returned 429, falling back to google/gemini-2.5-pro
[brcc] ↺ Fallback successful: using google/gemini-2.5-pro
To disable fallback:
brcc start --no-fallback┌─────────────────────────────────────────────────────────────┐
│ Claude Code │
│ (thinks it's talking to Anthropic) │
└──────────────────────┬──────────────────────────────────────┘
│ Anthropic Messages API format
▼
┌─────────────────────────────────────────────────────────────┐
│ brcc proxy (localhost:8402) │
│ │
│ 1. Receives request from Claude Code │
│ 2. Replaces model name (if --model set) │
│ 3. Signs x402 USDC payment with your wallet │
│ 4. Forwards to BlockRun API │
│ 5. Streams response back to Claude Code │
└──────────────────────┬──────────────────────────────────────┘
│ + x402 payment signature
▼
┌─────────────────────────────────────────────────────────────┐
│ BlockRun API (blockrun.ai or sol.blockrun.ai) │
│ │
│ Routes to: GPT-5 · Claude · Gemini · DeepSeek · Grok · │
│ NVIDIA · MiniMax · Moonshot · 50+ models │
└─────────────────────────────────────────────────────────────┘
Your private key stays on your machine. Only payment signatures are sent.
brcc uses USDC — a dollar-pegged stablecoin. No crypto volatility.
Base chain (default):
- Buy USDC on Coinbase
- Send to your brcc wallet address (shown in
brcc setup) - Make sure it's on Base network (not Ethereum mainnet)
Solana chain:
- Buy USDC on any exchange
- Send to your brcc Solana address
- Send on Solana network
Typical cost: $0.001–$0.05 per interaction. $5 lasts most developers a week.
Do I need an Anthropic account? No. brcc connects Claude Code to BlockRun instead of Anthropic.
Can I use non-Claude models? Yes. GPT-5, Gemini, DeepSeek, Grok, and 50+ others work through Claude Code via brcc.
- BlockRun — The AI gateway powering brcc
- npm package
- Roadmap
- Telegram
- GitHub Issues
Business Source License 1.1 — Free to use, modify, and deploy. Cannot be used to build a competing hosted service. Converts to MIT in 2036.