oztenbot - OpenClaw Global Hackathon: Building Autonomous Agentic Workflows
AI Tinkerers - Seattle
Hackathon Showcase

oztenbot

agent-meter is an open-source middleware providing per-agent budget enforcement and real-time usage tracking via three lines of code.

1 member

agent-meter: Budget Enforcement for the Agent Economy

Elevator Pitch

Agents are getting wallets, API keys, and autonomy — but nobody is metering them. agent-meter is an open-source usage metering layer that gives every agent a budget and every API a kill switch. Three lines of middleware. No code changes in your handlers.

Why This Matters

The agent economy has an accountability gap. Agents can call APIs, spend money, and consume resources — but there’s no standard way to track, limit, or bill for that usage. Today’s solutions are either “no limits” (dangerous) or “hard-coded rate limits” (one-size-fits-all). Neither works when agents have different trust tiers, budgets, and consumption patterns.

agent-meter fills this gap with:

  • Per-agent metering — Every request attributed to a specific agent identity
  • Budget enforcement via rate limits — Agents that exceed their budget get 429’d, not silently throttled
  • Real-time visibility — A live dashboard shows exactly who is consuming what, right now
  • Conversational access via MCP — Agents can query their own usage data through natural language

The Demo

What you’ll see:

  1. A real-time dashboard showing four agents hitting an API with different consumption patterns
  2. A well-behaved research-bot that stays within budget
  3. A rogue-bot that floods the API and gets rate-limited within seconds — red 429 entries appear in the live feed while other agents continue unaffected
  4. An MCP server that lets Claude query the same metering data conversationally — “what’s my usage summary?”

The punchline: Metering IS budget enforcement. The same system that tracks usage also enforces limits. And agents can introspect their own budgets through MCP — no dashboard required.

Run It

cd ~/agent-meter-server && bash demo/run.sh  

Architecture

┌─────────────┐     POST /v1/usage/batch     ┌──────────────────┐  
│  Agent SDK  │ ────────────────────────────▶ │  agent-meter     │  
│  (3 lines)  │ ◀──── 200 OK or 429 ──────── │  server (Rust)   │  
└─────────────┘                               │                  │  
                                              │  SQLite + rate   │  
┌─────────────┐     GET /v1/summary           │  limits per      │  
│  Dashboard  │ ────────────────────────────▶ │  (service,agent) │  
│  (browser)  │ ◀──── real-time JSON ──────── │                  │  
└─────────────┘                               └──────────────────┘  
                                                       ▲  
┌─────────────┐     MCP tools (13 total)               │  
│  Claude +   │ ───────────────────────────────────────┘  
│  MCP server │    "what's my usage summary?"  
└─────────────┘  

Links