TypeScript SDK

@memhq/sdk is the canonical client for MemHQ from a Node, edge, or browser-server agent. It ships full TypeScript types and works unchanged on Node 18+, Bun, Deno, and the major edge runtimes.

Install

npm install @memhq/sdk
# or
pnpm add @memhq/sdk
# or
yarn add @memhq/sdk

Configure

Set MEMHQ_API_KEY in your environment:

export MEMHQ_API_KEY="mhq_live_..."

The client picks it up automatically:

import { MemoryClient } from "@memhq/sdk";

const client = new MemoryClient();

Or pass the key explicitly:

const client = new MemoryClient({ apiKey: "mhq_live_..." });

Available constructor options:

Option	Default	Description
`apiKey`	`process.env.MEMHQ_API_KEY`	Project API key.
`baseUrl`	`https://api.memhq.ai`	Override for self-hosted control planes.
`timeout`	`30000` (ms)	Per-request timeout.
`maxRetries`	`3`	Retries on 429/5xx with exponential backoff.
`fetch`	global `fetch`	Inject a custom fetch (e.g. for proxy support).

Usage

The methods mirror the Memory API:

// Ingest
const episode = await client.add({
  userId: "user_42",
  messages: [{ role: "user", content: "I'm vegetarian." }],
});

// Search
const results = await client.search({
  userId: "user_42",
  query: "dietary preferences",
});

// Ask
const answer = await client.ask({
  userId: "user_42",
  question: "What should I order them?",
});
console.log(answer.text);

Streaming `ask`

ask supports server-sent events for token-by-token streaming:

const stream = await client.ask.stream({
  userId: "user_42",
  question: "Summarize their food preferences.",
});

for await (const chunk of stream) {
  process.stdout.write(chunk.delta);
}

Coming soon — full guide. Error classes, the dashboard-side client surface, and framework-specific recipes (Next.js, Hono, Vercel AI SDK) will be documented in the next pass.

TypeScript SDK

TypeScript SDK

Install

Configure

Usage

Streaming ask

On this page

Streaming `ask`