Persistent memory for LangChain

LangChain forgets everything between sessions. Connect it to Kit for AI and it saves what matters as it works — and recalls it in any later session, on any machine.

Why LangChain forgets

LangChain gives you memory classes, but you still have to run what's behind them — a vector store, an embedder, dedup logic, persistence. Point your chains and agents at a hosted memory API instead: two endpoints replace the infrastructure, and the store is shared beyond LangChain.

What you get

• remember / recall as tools the agent calls on its own — no manual note-keeping.
• Retrieval by meaning: "what did we decide about auth?" finds the JWT decision.
• One shared store: what this tool learns, Claude, Cursor, and your code can recall too.
• Near-duplicate detection and versioning keep months of memories clean.
• Same account ingests documents and URLs — memory and knowledge in one retrieval system.

Set it up in about 2 minutes

1. 1Create a free API key at kitforai.com/app/settings.
2. 2Wrap the REST endpoints (below) as a LangChain tool, or call them in your chain's pre/post hooks.
3. 3remember after meaningful turns; recall before generation and inject results into context.
4. 4The same memories are visible to MCP agents and any other client — no LangChain lock-in.

# Remember a fact (deduped automatically)
curl -X POST https://kitforai.com/api/v1/memory \
  -H "x-api-key: YOUR_API_KEY" -H "content-type: application/json" \
  -d '{"content":"We ship on Fridays; staging DB is read-only."}'

# Recall by meaning
curl "https://kitforai.com/api/v1/memory/search?q=deploy%20schedule" \
  -H "x-api-key: YOUR_API_KEY"

FAQ

Why not just use LangChain's built-in memory classes?

They're interfaces — you still operate the vector DB and embedder behind them, and the memory stays inside your app. Kit for AI hosts the store, handles retrieval quality and dedup, and shares it across every tool you use.

Does LangChain save memories automatically?

Yes — once connected, the agent has remember and recall as callable tools and uses them on its own: it saves decisions and preferences mid-conversation and retrieves them when they're relevant. You can also tell it explicitly to remember something.

Is my data private?

Memories are owner-scoped to your account, served over HTTPS, with encryption at rest for stored content and an audit trail of what agents wrote. You can review, edit, and delete every memory from the dashboard.

Can my other AI tools share this memory?

Yes — that's the point. The same store is reachable from every MCP client and over REST, so what LangChain learns, Claude, Cursor, and your own code can recall too.

Is it free?

There's a free plan — an account and API key cost nothing, and setup takes about two minutes. Paid plans add higher quotas and team features.