One ingest endpoint. Six ways to send.
Use a native SDK, point your LLM gateway at us, ship OpenTelemetry GenAI traces, connect a SaaS receiver, run our MCP server, or hand us your CASB feed. We meet your stack where it is.
pip install agentics
export AGENTICS_API_KEY=ak_...
from agentics import Agentics
client = Agentics()
with client.track(
agent_id="support-bot",
foundation_model="anthropic/claude-opus-4-6",
model_version="2026-04-22",
scope=["read:tickets","write:reply"],
) as task:
reply = my_llm_call(...)
task.set_outputs(reply)
task.set_outcome("success")
npm i @agentics/sdk
export AGENTICS_API_KEY=ak_...
import { Agentics } from "@agentics/sdk";
const client = new Agentics();
await client.track(
{ agentId: "support-bot",
foundationModel: "anthropic/claude-opus-4-6",
modelVersion: "2026-04-22",
scope: ["read:tickets","write:reply"] },
async (task) => {
const reply = await myLlmCall(...);
task.setOutputs(reply).setOutcome("success");
});
Bearer-auth every request with an API key (ak_...) from /console/keys or a JWT session cookie.
| Endpoint | Method | Purpose |
|---|---|---|
/api/sdk/ingest | POST | Record a receipt (single or {batch:[...]}) |
/api/gateway/proxy | POST | Accept LLM-gateway callbacks (LiteLLM/Portkey/Helicone/etc.) |
/api/otel/ingest | POST | OpenTelemetry GenAI OTLP/JSON |
/api/receivers/:vendor | POST | Native SaaS receivers (Claude Enterprise, OpenAI Enterprise, Copilot, etc.) |
/api/casb/:vendor | POST | CASB shadow-AI events (Netskope, Zscaler, Defender, etc.) |
/api/mcp/manifest | GET | MCP server discovery |
/api/mcp/jsonrpc | POST | MCP server entrypoint |
/api/agents-e | GET | List enterprise agents |
/api/policies/evaluate | POST | Evaluate a policy decision |
/api/risk | GET/POST | AI Risk Register CRUD |
/api/incidents | GET/POST | Incidents with timeline + postmortem |
/api/approvals | GET/POST | Multi-step approvals |
/api/evidence | GET/POST | Evidence packs |
/api/redteam | GET/POST | Red team runs + findings |
/api/drift | GET | Per-agent drift detection |
/api/bias | GET | Bias signals per agent |
/api/aibom/:agent_id | GET | CycloneDX 1.6 AI-BOM |
/api/tenants/me | GET/PATCH | Tenant settings |
/api/api-keys | GET/POST/DELETE | API key management |
/api/webhooks | GET/POST/DELETE | Outbound webhooks (9 event types) |
/api/exports | POST | Bulk evidence/receipt export |
/api/litigation/holds | GET/POST | Litigation holds |
/api/retention | GET/PATCH | Retention windows |
/api/rtbf | POST | Right-to-be-forgotten requests |
/api/trust/* | GET | Public trust-center endpoints |
/api/intelligence/* | GET | Industry packs, benchmarks (k-anon 5), threat feed |
/api/insurance/* | GET/POST | Partners + referrals |
/api/operators/* | GET/POST | Operator marketplace + engagements |
/api/console/summary | GET | Per-tenant dashboard rollup |
Point any MCP-capable client at https://agentics.you/api/mcp/manifest. Tools exposed: search_agents, get_agent, record_receipt, list_certifications. Every call records a receipt for the calling agent so MCP usage itself is auditable.
Subscribe at /console/webhooks. Events: receipt.recorded, violation.detected, policy.blocked, incident.opened, incident.closed, certification.granted, certification.revoked, agent.created, tenant.updated.
Each delivery is signed: X-Agentics-Signature: t=…,v1=HMAC_SHA256(secret,t+"."+body).