AI Memory for Chatbots, Agents, and LLM Apps

Your AI forgets everything. CLAIV fixes that.

If you're building with OpenAI, Claude, or LangChain — you've seen this:

• Your chatbot forgets users between sessions
• Your AI agent loses context after a few steps
• Your app relies on fragile chat history or prompt stuffing
• Your RAG setup is inconsistent and unreliable

👉 This is not a model problem — it's a memory problem

CLAIV is a drop-in memory API for AI applications.

It gives your AI:

• 🧠 Persistent user memory across conversations
• 📚 Reliable document understanding (not just chunk retrieval)
• ⏱️ Temporal awareness — what happened and when
• 🎯 Context injection ready for LLM prompts
• 🧹 Controlled forgetting — GDPR-safe deletion

❌ Without memory

User: My name is Alex and I run a fitness business.
...later...
AI: How can I help you today?

✅ With CLAIV

AI: How is your fitness business going, Alex?

1. Send events to POST /v6/ingest
2. CLAIV extracts structured memory asynchronously
3. Call POST /v6/recall before each LLM response
4. Inject llm_context.text into your system prompt

That's it.

pip install claiv-memory

fromclaivimportClaivClientclient=ClaivClient(api_key="your_key")
client.ingest({
"user_id": "user_123",
"conversation_id": "conv_abc",
"type": "message",
"role": "user",
"content": "I'm building an AI agent for compliance review."
})

context=client.recall({
"user_id": "user_123",
"conversation_id": "conv_abc",
"query": "What is the user building?"
})
print(context["llm_context"]["text"])

System prompt:
User memory:
{{ llm_context.text }}

JavaScript / TypeScript

npm install @claiv/memory

import{ClaivClient}from'@claiv/memory';constclient=newClaivClient({apiKey: 'your_key'});awaitclient.ingest({user_id: 'user_123',conversation_id: 'conv_abc',type: 'message',role: 'user',content: "I'm building an AI agent for compliance review.",});constcontext=awaitclient.recall({user_id: 'user_123',conversation_id: 'conv_abc',query: 'What is the user building?',});

Upload documents and CLAIV will:

• Parse structure into sections and spans
• Embed content with pgvector
• Retrieve relevant sections automatically
• Inject them into your LLM context

doc=client.upload_document({
"user_id": "user_123",
"project_id": "proj_abc",
"document_name": "Product Manual",
"content": open("manual.md").read(),
})
print(f"Indexed {doc['spans_created']} spans across {len(doc['sections'])} sections")

Supports:

• Knowledge copilots over internal documentation
• Compliance and research tools
• Support assistants that reason over policies and manuals

POST /v6/ingest → Store memory eventsPOST /v6/recall → Retrieve contextPOST /v6/documents → Upload documentsPOST /v6/forget → Delete memory

Full reference → claiv.io/docs

LoCoMo 10-dialogue J-score: 75.0%

LoCoMo is the standard benchmark for long-context conversational memory in production AI systems.

• AI chatbots with persistent memory
• AI agents with multi-step workflows
• SaaS apps with per-user context
• Customer support systems
• Internal knowledge copilots
• Research and compliance tools

• Ingest is async — memory may not appear instantly
• Always include user_id and conversation_id
• Inject llm_context.text into your system prompt

• API key authentication with tenant isolation
• Strict schema validation on all endpoints
• Scoped deletion via POST /v6/forget

👉 Get API key:claiv.io

👉 Docs:claiv.io/docs