runtimeai 1.0.0

RuntimeAI Platform SDK — Govern, secure, and observe AI agents at scale

runtimeai

Official Python SDK for the RuntimeAI Control Plane — govern, secure, and observe AI agents at scale.

Installation

pip install runtimeai

Quick Start

from runtimeai import RuntimeAI

client = RuntimeAI(
    "https://api.rt19.runtimeai.io",
    api_key="your-api-key",
)

# Register an AI agent
agent = client.agents.register(
    name="contract-analyzer",
    type="langchain",
    owner="[e-mail]",
    risk_level="medium",
)
print(f"Agent registered: {agent['id']}")

# Discover shadow AI usage
findings = client.discovery.list_findings()
print(f"Found {len(findings.get('findings', []))} shadow AI instances")

# Track costs
client.finops.report_usage(
    agent_id=agent["id"],
    model="gpt-4",
    provider="openai",
    input_tokens=1500,
    output_tokens=800,
    cost_usd=0.045,
)

# Check compliance posture
frameworks = client.compliance.list_frameworks()
for fw in frameworks.get("frameworks", []):
    print(f"{fw['name']}: {fw['compliance_pct']}% compliant")

Authentication

API Key

client = RuntimeAI(
    "https://api.your-pod.runtimeai.io",
    api_key="your-api-key",  # X-API-Key header
)

JWT Token

client = RuntimeAI(
    "https://api.your-pod.runtimeai.io",
    auth_token="your-jwt-token",  # Authorization: Bearer header
    tenant_id="your-tenant-id",   # X-Tenant-ID header
)

Available Clients

Client	Methods
client.agents	register, list, get, update, kill_switch, get_trust_score, heartbeat
client.discovery	list_findings, triage_finding, list_scanners, configure_scanner, ingest
client.finops	report_usage, get_usage_summary, list_budgets, set_budget, get_cost_attribution
client.policies	list_guardrails, create_guardrail, list_sod_rules, list_conditional_access, evaluate
client.mcp	list_tools, register_tool, invoke_tool, list_connections, get_topology
client.compliance	list_frameworks, list_gaps, upload_evidence, list_risk_scores
client.audit	list, search, get

Configuration

Option	Type	Default	Description
api_url	str	—	Required. Control Plane API URL
api_key	str	""	API key (X-API-Key header)
auth_token	str	""	JWT token (Bearer header)
tenant_id	str	""	Tenant ID for multi-tenant ops
timeout	int	30	HTTP timeout in seconds
max_retries	int	2	Max retries on 5xx errors

Error Handling

from runtimeai import RuntimeAI, RuntimeAIError

try:
    agent = client.agents.get("nonexistent-id")
except RuntimeAIError as e:
    print(f"API Error: {e.status_code} {e.code} — {e}")

Requirements

• Python >= 3.9
• Zero runtime dependencies (stdlib urllib only)

License

MIT © RuntimeAI