Overview

Memory infrastructure for AI applications. Store, search, and generate answers — all through a single API.

Build AI that remembers

MemoryKit gives your AI application persistent memory. Store any content and retrieve it with hybrid search — in three lines of code.

import { MemoryKit } from "memorykit";
const mk = new MemoryKit({ apiKey: "ctx_..." });
 
// Store anything
await mk.memories.create({ content: "Our refund policy is 30 days." });
 
// Search your data
const results = await mk.memories.search({ query: "refund policy" });
console.log(results.results[0].content);
// → "Our refund policy is 30 days."

Start building

Store

Ingest text, PDFs, DOCX, and structured data. MemoryKit chunks, embeds, and indexes automatically.

Search

Hybrid retrieval combines vector similarity, full-text search, and reranking in one call.

Retrieve

Get ranked results with relevance scores and feed them into your own LLM or application logic.

What developers build

headsetSupport chatbot

Replace static FAQs with a chatbot that answers from your docs, policies, and past tickets.

book-openKnowledge base

Give your team instant answers from scattered docs, Notion pages, and Confluence wikis.

userPersonal assistant

Build an assistant that remembers each user's preferences and history.

file-textDocument Q&A

Upload a PDF and start asking questions about it immediately.

Base URL

https://api.memorykit.io/v1

SDKs

Full type safety. Node.js, Bun, Deno, edge runtimes. Zero dependencies.

Sync and async clients. Python 3.8+. FastAPI, Django, Flask.

Native async/await. iOS 15+, macOS 12+, watchOS, tvOS.

Coroutines and Flow. Android 21+, JVM 8+.

Full setup and memory management from the terminal. Agent-friendly.

Next steps

Store and search your first memory in 30 seconds.

Create an API key and secure your requests.

How memories are processed, stored, and retrieved.

Next

Quickstart

On this page

Build AI that remembers

What developers build