memorysync-langchain

MemorySync integration for LangChain.js: persistent chat message history, structured memory tools for agents, and prompt-ready recall context. Works with @langchain/core 0.3.x and 1.x.

npm install memorysync-langchain @langchain/core

Set MEMORYSYNC_API_KEY (create a key at memorysync.io, or run npx memorysync-cli init).

Chat message history

MemorySyncChatMessageHistory extends BaseListChatMessageHistory, so it plugs straight into RunnableWithMessageHistory:

import { RunnableWithMessageHistory } from "@langchain/core/runnables";
import { MemorySyncChatMessageHistory } from "memorysync-langchain";

const chainWithHistory = new RunnableWithMessageHistory({
  runnable: chain,
  getMessageHistory: (sessionId) => new MemorySyncChatMessageHistory({ sessionId }),
  inputMessagesKey: "input",
  historyMessagesKey: "history",
});

await chainWithHistory.invoke(
  { input: "My name is Ada." },
  { configurable: { sessionId: "thread-42" } },
);

Turns are stored verbatim through MemorySync's episodic ingestion — no extraction gates, so short turns like "yes" survive — and the full LangChain message (tool calls, kwargs) is serialised into metadata for exact reconstruction. Writes are idempotent: a retried write is recognised, never duplicated. Plain-text transcripts use the same stored shape as the Python integration, so both SDKs can read them.

const history = new MemorySyncChatMessageHistory({
  sessionId: "thread-42",
  userId: "customer-7",   // share memory across a user's sessions
  maxMessages: 30,        // cap the transcript tail handed to the LLM
});

Agent tools

import { createMemorySyncTools } from "memorysync-langchain";

const tools = createMemorySyncTools({ endUserId: "customer-7" });
// langchain 1.x
import { createAgent } from "langchain";
const agent = createAgent({ model, tools });

Returns add_memory, search_memory, list_memories, update_memory, and delete_memory. Tools return readable strings and never throw, so a memory failure cannot abort an agent run. Pass readOnly: true to hand an agent only search_memory and list_memories.

add_memory derives an idempotency key from the content — an agent that repeats itself gets "already stored", not a duplicate.

Recall context

import { MemorySyncContextProvider } from "memorysync-langchain";

const provider = new MemorySyncContextProvider({ userId: "customer-7" });
const context = await provider.getContext("What should I cook tonight?");

Returns a grouped, type-labelled context block built by MemorySync's hierarchical retrieval, ready to inject into a system prompt. Empty string — not an exception — when the user has no relevant memories yet.

Documentation

Full guide: docs.memorysync.io/integrations/langchain