Security & Guardrails12.8ms Reflex$0 / MIT License
17,400 active downloads
Token Budget Limiter
Hard rate limit and session spending enforcement filter
Author: CostGuard AI
GitHub Repository →jevproxy // token-budget-limiter (reflex kernel)
$npx jevproxy run token-limiter
JevProxy Intercept Resolved(Mode: DIRECT_REFLEX)
12.8ms|Cost: $0.0001|0 reasoning tokens burned
STDOUT • Tool Call Output:Status: 200 OK
agent.dispatch(check_token_budget)
payload: { "session_id": { "type": "string", "description": "User session identifier" }, "estimated_to...
✔ Decision short-circuited in 12.8ms without roundtrip to frontier LLM.
Upstream token bill saved: $0.0090 on this turn.
Token Budget Limiter
Compatible with Cursor, Claude Code, Windsurf, OpenCode
TRADITIONAL LLM CALL:UNOPTIMIZED
• Median Latency: 950 ms
• Cost per Turn: $0.0090
• Mode: Full KV Cache Reload & TTFT Prefill
JEVPROXY REFLEX KERNEL:77x FASTER
• Median Latency: 12.8 ms
• Cost per Turn: $0.0001
• Accuracy: 100.0% deterministic
1-Click CLI Execution
Run this tool accelerated through the JevProxy gateway without manual wiring:
npx jevproxy run token-limiterOpenAI / Anthropic Tool Schema
JSON SpecificationPaste this schema into your agent tools definition or Cursor extensions:
{
"type": "function",
"function": {
"name": "check_token_budget",
"description": "Verify session balance before allowing prompt execution",
"parameters": {
"type": "object",
"properties": {
"session_id": {
"type": "string",
"description": "User session identifier"
},
"estimated_tokens": {
"type": "number",
"description": "Anticipated prompt token length"
}
},
"required": [
"session_id",
"estimated_tokens"
]
}
}
}Technical Architecture & Usage
Preventing Runaway Agent Bills
A loop of recursive tool calls can easily burn $50 in 10 minutes. Token Budget Limiter provides hardware-level budget enforcement.
Accelerate Token Budget Limiter with JevProxy
Get 5,000,000 free decision tokens. Eliminate the 3-second tool freeze in Cursor and Claude Code in under 60 seconds.