DEVELOPERS

Build agents that understand the world

Stop rebuilding context pipelines, memory systems, entity matching, connectors, and state management. Four primitives, one API.

import { PointOfData } from "@pointofdata/sdk";

const pod = new PointOfData({ apiKey: process.env.POD_KEY });

// stream real-world signals in
await pod.signals.ingest({
  source: "shopify",
  entity: { type: "order", id: "48812" },
  event:  "shipment.delayed",
  at:     new Date(),
  payload: { carrier: "SF", days: 3 }
});

The workflow

STEP 01
Connect
Bring in real-world signals from the systems you already run.
pod.signals.ingest()
STEP 02
Understand
Entities, relationships, and history resolve continuously.
pod.understand()
STEP 03
Build
Agents reason over situations, not retrieved chunks.
pod.context.forAgent()
STEP 04
Act
Actions run inside policy, and every one is recorded.
pod.actions.record()

Documentation

Full library — including explainers and benchmarks — in Resources.

Quickstart
First signal to first situation brief in ten minutes.
→
Core concepts
Signals, entities, relationships, state, history.
→
Entity resolution
Matching strategies, confidence, and overrides.
→
Situation briefs
Shaping context for a specific agent and task.
→
Policy & permissions
Scoping what an agent may know and do.
→
API reference
Every endpoint, every field, every error.
→

SDKs & integrations

typescript
v2.4.0
python
v2.4.0
go
v1.9.2
rust
v0.8.1
java
v1.3.0
ruby
v1.1.4
cli
v3.0.1
terraform
v0.6.0
MCP server
Expose the understanding layer to any MCP-compatible agent runtime with one command.
npx @pointofdata/mcp connect

Start with a sandbox project and 1M free signals.

Get API access →