Installation | Hivemind docs

npm

npm install @get-hivemind/sdk

bun

bun add @get-hivemind/sdk

import { Hivemind } from "@get-hivemind/sdk";

const hm = new Hivemind({ apiKey: process.env.HIVEMIND_API_KEY! });

await hm.remember({ content: "We cap production deploys at 3 a month" });

const { payload } = await hm.recall({ q: "how often can we deploy?" });

Create a key under Connected tools in the dashboard. Treat it like a password, it grants full access to your memory.

Options

	Default
`apiKey`	required	From the dashboard
`sourceRef`	none	Default origin for calls that do not name one, or `withOrigin()`
`maxRetries`	`2`	`0` disables retrying
`timeoutMs`	`30000`	Per attempt
`baseUrl`	hosted	Point at a different deployment
`fetch`	global	Inject your own, for tests or a proxy

Asking from one origin

const acme = hm.withOrigin("acme");
await acme.remember({ content: "Their staging runs Postgres 16" });

Everything on the derived client names that origin. Searches still cover everything owned; results from the origin are preferred, never exclusive. The original is unchanged.

Retries

Transient failures, 408, 409, 429, and 5xx, are retried with exponential backoff and jitter. Client errors are not: retrying a 400 cannot succeed, and retrying a quota error only delays the message you need to read.