Integrate constitutional AI governance in minutes, not months.
A single API call evaluates any AI agent action against all 792 AGS v2.1 governance dimensions.
import agentshield client = agentshield.Client(api_key="your-api-key") # Evaluate an agent action against AGS v2.1 result = client.evaluate( agent_id="agent-001", action={ "type": "data_access", "target": "customer_records", "intent": "generate quarterly report" } ) print(result.decision) # "ALLOW" or "BLOCK" print(result.score) # 0.0 — 1.0 governance score print(result.dimensions) # 792 dimension breakdown
import { AgentShield } from '@agentshield/sdk'; const client = new AgentShield({ apiKey: 'your-api-key' }); // Evaluate an agent action against AGS v2.1 const result = await client.evaluate({ agentId: 'agent-001', action: { type: 'data_access', target: 'customer_records', intent: 'generate quarterly report' } }); console.log(result.decision); // "ALLOW" or "BLOCK" console.log(result.score); // 0.0 — 1.0 governance score console.log(result.dimensions); // 792 dimension breakdown
| Method | Endpoint | Description |
|---|---|---|
| POST | /v1/evaluate | Evaluate an agent action against all 792 AGS v2.1 dimensions. Returns governance decision, score, and full dimension breakdown. |
| GET | /v1/agent/{id}/score | Retrieve the current aggregate governance score for a specific agent, including historical trend data. |
| GET | /v1/lockdown/status | Check the current lockdown status across all governed agents. Returns active lockdowns and trigger conditions. |
| GET | /v1/dimensions | List all 792 AGS v2.1 governance dimensions with descriptions, weights, and current evaluation criteria. |
SDKs across six languages and platforms. Python and TypeScript are full-featured (type safety, retries, streaming); Go, Rust, and Node.js provide core evaluation clients.
The AGS v2.1 standard is maintained by AgentGoverning, the independent standards body for AI agent governance.
agentgoverning.com →