agent-ready-client 0.3.1

Official Python client SDK for the Agent Ready API — scan any URL for AI agent-readability (Vercel Agent Readability Spec, llmstxt.org, agent-protocol manifests).

agent-ready-client (Python SDK)

Official Python client SDK for the Agent Ready API — scan any public URL for AI agent-readability against the Vercel Agent Readability Spec, the llmstxt.org standard, and agent-protocol manifests (MCP server cards, A2A, agents.json, agent-permissions.json, UCP, x402, NLWeb).

Zero runtime dependencies — pure standard library (urllib). Python 3.8+.

Prefer the terminal? Use the agent-ready-scanner CLI. Working in JS/TS? See agent-ready-client on npm. This package is for calling the API from your own Python code.

Install

pip install agent-ready-client

Quick start

import os
from agent_ready import AgentReady

ar = AgentReady(api_key=os.environ["AGENT_READY_API_KEY"])

scan = ar.scan("https://example.com")          # start + poll to completion
print(scan["vercelScore"], scan["vercelRating"])  # 96 "excellent"

for check in scan["siteChecks"]:
    if check["status"] == "fail":
        print(check["checkId"], check["name"], "->", check["howToFix"])

Authentication

scan, start_scan, get_scan, and list_scans require a Pro API key (issue one at https://agent-ready.dev/dashboard/api-keys). Pass it explicitly or set AGENT_READY_API_KEY in the environment:

ar = AgentReady()                       # reads AGENT_READY_API_KEY
ar = AgentReady(api_key="ar_live_...")  # or pass it in

ask is public and needs no key.

API

AgentReady(api_key=None, base_url="https://agent-ready.dev", timeout=30.0)

Method	Returns	Notes
scan(url, page_limit=None, poll_interval=2.0, timeout=120.0)	Scan	Start and poll to completion.
start_scan(url, page_limit=None)	StartScanResponse	Queue only; returns the id.
get_scan(id)	Scan	Fetch a scan (running or finished).
list_scans(limit=None, cursor=None)	ScanListResponse	Your scans, newest first.
ask(query, item_type=None, mode=None)	dict	NLWeb doc search. No key required.
scan_mcp(endpoint)	McpScanResponse	Grade a live MCP server (tools/resources/prompts). No key required.

Fire-and-forget + poll later

started = ar.start_scan("https://example.com", page_limit=25)
# ...later...
scan = ar.get_scan(started["id"])
if scan["status"] == "completed":
    print(scan["vercelScore"])

Errors

Every failure raises ApiError with a stable code and (when from a response) an HTTP status:

from agent_ready import ApiError

try:
    ar.scan("https://example.com")
except ApiError as err:
    if err.code == "rate_limited":
        ...  # back off and retry

Common codes: missing_api_key, unauthorized, subscription_required, rate_limited, invalid_request, timeout, network_error.

Links

• API docs & OpenAPI 3.1 spec: https://agent-ready.dev/docs/api · https://agent-ready.dev/api/v1/openapi.json
• Methodology (all checks): https://agent-ready.dev/methodology
• JS/TS SDK: https://www.npmjs.com/package/agent-ready-client · CLI: https://www.npmjs.com/package/agent-ready-scanner

License

MIT © Agent Ready