Your agent hit the API rate limit and started failing silently. AXME enforces rate and cost limits at the gateway - before the damage is done.
AI agents make API calls in loops. Without centralized rate control, one runaway agent can exhaust your API quota, burn through your budget, and take down the whole system. AXME cost policies let you set hard limits per agent: max intents per day, max intents per hour, max cost per day in USD.
Alpha - Built with AXME (AXP Intent Protocol). cloud.axme.ai - contact@axme.ai
Agent "order-processor" in production:
09:00 - processes 50 orders (normal)
09:15 - retry loop hits a flaky upstream API
09:16 - 200 retries/minute (exponential backoff broken)
09:30 - 5,000 API calls made, $47 in LLM costs
09:45 - 12,000 API calls, $130 in costs, rate-limited by upstream
10:00 - all other agents sharing the API key are now blocked
11:00 - someone finally notices
What should have happened:
09:00 - processes 50 orders (normal)
09:15 - retry loop starts
09:16 - AXME: 200 intents/hour limit reached. 429 returned.
09:16 - Agent stops. Alert fires. $0.80 spent, not $130.
import httpx, os
api_key = os.environ["AXME_API_KEY"]
base = "https://cloud.axme.ai"
headers = {"x-api-key": api_key}
# Set a cost policy on the agent
address = "agent://myorg/production/order-processor"
httpx.put(
f"{base}/v1/mesh/agents/{address}/policies/cost",
headers=headers,
json={
"max_intents_per_hour": 200,
"max_intents_per_day": 2000,
"max_cost_per_day_usd": 10.00,
"action_on_breach": "block", # or "alert", "require_approval"
},
)When the agent exceeds any limit, AXME returns 429 Too Many Requests with a Retry-After header. The agent stops burning money immediately.
pip install axme httpx
export AXME_API_KEY="your-key" # Get one: axme login# set_policy.py
import httpx
import os
api_key = os.environ["AXME_API_KEY"]
base_url = os.environ.get("AXME_BASE_URL", "https://cloud.axme.ai")
headers = {"x-api-key": api_key}
agent_address = "agent://myorg/production/order-processor"
response = httpx.put(
f"{base_url}/v1/mesh/agents/{agent_address}/policies/cost",
headers=headers,
json={
"max_intents_per_hour": 200,
"max_intents_per_day": 2000,
"max_cost_per_day_usd": 10.00,
"action_on_breach": "block",
},
)
print(f"Policy set: {response.json()}")# send_until_limited.py
from axme import AxmeClient, AxmeClientConfig
import os
client = AxmeClient(AxmeClientConfig(api_key=os.environ["AXME_API_KEY"]))
for i in range(250):
try:
intent_id = client.send_intent({
"intent_type": "intent.orders.process.v1",
"to_agent": "agent://myorg/production/order-processor",
"payload": {"order_id": f"ORD-{i:04d}"},
})
print(f"[{i}] Sent: {intent_id}")
except Exception as e:
print(f"[{i}] Blocked: {e}")
# Agent hit the rate limit - this is expected after 200 intents
break# check_usage.py
import httpx
import os
api_key = os.environ["AXME_API_KEY"]
base_url = os.environ.get("AXME_BASE_URL", "https://cloud.axme.ai")
headers = {"x-api-key": api_key}
agent_address = "agent://myorg/production/order-processor"
response = httpx.get(
f"{base_url}/v1/mesh/agents/{agent_address}/policies/cost",
headers=headers,
)
policy = response.json()
print(f"Policy: {policy}")import redis
import time
r = redis.Redis()
def rate_limited_call(agent_id, func, *args):
key = f"rate:{agent_id}:{time.strftime('%Y%m%d%H')}"
count = r.incr(key)
r.expire(key, 3600)
if count > 200:
raise Exception("Rate limit exceeded")
daily_key = f"rate:{agent_id}:{time.strftime('%Y%m%d')}"
daily_count = r.incr(daily_key)
r.expire(daily_key, 86400)
if daily_count > 2000:
raise Exception("Daily limit exceeded")
# Still need: cost tracking, USD limits, per-agent policies,
# alerting, approval workflows, audit trail, dashboard...
return func(*args)httpx.put(
f"{base_url}/v1/mesh/agents/{address}/policies/cost",
headers={"x-api-key": api_key},
json={
"max_intents_per_hour": 200,
"max_intents_per_day": 2000,
"max_cost_per_day_usd": 10.00,
"action_on_breach": "block",
},
)Rate limiting, cost caps, breach actions, audit trail, and dashboard - all built in.
cost policy check
+-----------+ send_intent() +----------------+ deliver +-----------+
| | -------------> | | ---------> | |
| Initiator | | AXME Cloud | | Agent |
| | | (gateway) | | |
+-----------+ | | +-----------+
| counters: |
| - hourly: 197 |
| - daily: 1842 |
| - cost: $7.20 |
| |
| limit hit? |
| YES -> 429 |
| NO -> deliver|
+----------------+
|
mesh.axme.ai
(usage dashboard)
View real-time usage counters, policy status, and breach history for every agent on your mesh:
| Field | Type | Description |
|---|---|---|
max_intents_per_hour |
int |
Maximum intents allowed per rolling hour |
max_intents_per_day |
int |
Maximum intents allowed per calendar day |
max_cost_per_day_usd |
float |
Maximum USD cost per calendar day |
action_on_breach |
string |
"block" (429), "alert" (allow + notify), "require_approval" (hold for human) |
curl -fsSL https://raw.githubusercontent.com/AxmeAI/axme-cli/main/install.sh | sh
axme login
pip install axme httpxpython set_policy.py
axme scenarios apply scenario.jsonAXME_API_KEY=<agent-key> python agent.pypython send_until_limited.pypython check_usage.py
# Shows current counters vs policy limits- AXME - project overview
- AI Agent Cost Monitoring - track spend per agent
- AI Agent Kill Switch - emergency stop
- AI Agent Health Monitoring - liveness and readiness
- AXME Examples - 20+ runnable examples
MIT - see LICENSE.
Built with AXME (AXP Intent Protocol).