Your first context query

Once your agent is connected, querying context is the same operation under the hood — the query-context tool — whether you call it over MCP, REST, or an SDK. Here is the same first query, three ways.

Replace <YOUR_API_KEY> with your fab_sk_… key and ws_<your_project> with your workspace ID. Use the URLs shown in your console.

MCP

With Fabric installed as an MCP server (see the per-client guides), just ask your agent a question that depends on your data:

“What are the top accounts at churn risk?”

The agent calls the query-context tool; Fabric traverses your context graph and returns the ranked entities and relationships. Nothing else to do — the result is grounded, scope-pinned, and audited.

REST

Terminal

curl https://api.wexa.ai/v1/query-context \
  -H "Authorization: Bearer <YOUR_API_KEY>" \
  -H "X-Wexa-Workspace: ws_<your_project>" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "top accounts at churn risk",
    "limit": 10
  }'

See Query context for the full request and response shape.

Python

agent.py

# pip install wexa
from wexa import Wexa

wexa = Wexa(api_key="<YOUR_API_KEY>")

context = wexa.context.query(
    workspace="ws_<your_project>",
    query="top accounts at churn risk",
)
print(context.entities)

See the Python SDK guide.

Node

agent.ts

// npm install @wexa/sdk
import { Wexa } from "@wexa/sdk";

const wexa = new Wexa({ apiKey: process.env.WEXA_API_KEY });

const context = await wexa.context.query({
  workspace: "ws_<your_project>",
  query: "top accounts at churn risk",
});
console.log(context.entities);

See the TypeScript / Node SDK guide.

What happened

Your query ran through Fabric’s governed pipeline: it was authenticated, scope-pinned to your workspace, validated, policy-checked, executed by graph traversal over your context graph, post-processed (PII redaction), and recorded in the audit log. The response carries a lifecycle_id you can use to trace and audit the call.

Because retrieval is graph traversal — not a full-corpus dump into the prompt — the result is the connected neighborhood your query actually needs, which keeps token cost low.

Next steps

• The Fabric MCP tools — query-context, create-ontology, save-context, docs.
• Simple vs Advanced mode — when writes and ontology proposals come into play.