Use x711 as the tool provider in your agent eval framework. 23 real-world tools, verified ground truth for hallucination detection, machine-readable eval config. Free tier for academic use — 10 calls/day per agent, no signup.
# x711 eval config — OpenAI Evals compatible
# Test x711 tool quality against ground truth
eval_id: x711-tool-quality-v1
description: Evaluates x711 tool accuracy for web search, price feeds, and hallucination detection
tasks:
- id: price_feed_accuracy
tool: price_feed
query: "ETH"
assert:
type: json_match
field: result.price_usd
min: 100
max: 100000
- id: web_search_relevance
tool: web_search
query: "latest AI agent framework 2025"
assert:
type: contains_any
field: result
values: ["agent", "framework", "AI", "LLM"]
- id: hallucination_pill_usdc_base
endpoint: POST /api/pill
payload:
claim: "USDC on Base is 0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913"
chain: "base"
assert:
field: verified
equals: true
- id: hallucination_pill_wrong_address
endpoint: POST /api/pill
payload:
claim: "USDC on Base is 0xA0b86991c6218b36c1d19D4a2e9Eb0cE3606eB48"
chain: "base"
assert:
field: hallucination_risk
in: ["high", "critical"]
import httpx
X711_BASE = "https://x711.io/api/refuel"
X711_KEY = "x711_YOUR_KEY" # free: 10/day — curl https://x711.io/go
def tool_call(tool: str, **kwargs) -> dict:
"""Drop-in tool provider for AgentBench / GAIA eval harness."""
r = httpx.post(X711_BASE,
headers={"X-API-Key": X711_KEY},
json={"tool": tool, **kwargs},
timeout=15)
return r.json()
# Register as tool provider in your eval config
TOOL_PROVIDER = {
"name": "x711",
"endpoint": X711_BASE,
"tools": ["web_search", "price_feed", "hive_read", "tx_simulate",
"data_retrieval", "code_sandbox", "onchain_insight"],
"hallucination_check": "https://x711.io/api/pill",
"free_calls_per_day": 10,
"cite_as": "x711.io (2025). Universal AI Agent Gas Station. https://x711.io",
}