Meko for LangChain
Meko gives your LangChain agents memory. A datapack carries what every chain learned, searchable before each run, by this agent and every other one that uses your datapack.
Over MCP, Meko provides LangChain with:
Agent-scoped memory
What your agent learns persists beyond the session and is readable via every AI tool you connect to.
A shared knowledge base
Documents and promoted conclusions are searchable by all agents and teammates with access to the datapack.
A trace of every read and write
Each conversation is stored with its reasoning, so you can see what an agent did and why.
Skip the stack you would wire yourself
One datapack replaces the vector store, memory add-on, and the sharing convention (reached through LangChain’s own first-party MCP adapter), so your dependency list stays as it is.
Operations covering memory, conversations, knowledge base, datapacks, artifacts, and observability are available for you. Your agent calls them mid-reasoning, and your code calls the same endpoint directly when an operation must happen every time.
How Meko Benefits LangChain
Give a LangGraph agent durable memory
An MCP library like LangChain’s own langchain-mcp-adapters can load all of Meko’s MCP tools as standard BaseTools over remote HTTP.
Recall before reasoning
Call the memory and knowledge base search tools before handing context to the agent, so recall runs on every request. The agent keeps the full toolset for mid-reasoning calls.
Save the conclusions that the run should keep
Saving a memory is an explicit call, stored with its reasoning and linked to related entities. Your code decides what persists.
Share the store beyond LangChain
The datapack a LangGraph app writes to is the same one Claude Code, Cursor, or a Strands agent reads from, tagged by agent_id.
Connecting Meko in LangChain
1. Install the adapter: pip install langchain-mcp-adapters
2. Register Meko as a remote server:
import asyncio
from langchain_mcp_adapters.client import MultiServerMCPClient
from langgraph.prebuilt import create_react_agent
mcp_client = MultiServerMCPClient({
"meko": {
"url": "https://mcp.mekodata.ai/mcp",
"transport": "streamable_http",
"headers": {
"Authorization": "Bearer <your-api-key>",
"User-Agent": "meko-langchain/1.0",
},
}
})
tools = asyncio.run(mcp_client.get_tools())
agent = create_react_agent("openai:gpt-4o-mini", tools)