Why customer memory matters
Every time a customer contacts support, they start from zero. The agent doesn't know about last week's issue, the preferences they mentioned, or the decision they made.
ConvoMem fixes this by providing a memory layer for conversational AI. It captures, stores, and retrieves customer context automatically.
How it works
- Capture — Every conversation is processed through our extraction pipeline
- Store — Facts are categorized, scored for durability, and stored with metadata
- Retrieve — When a new conversation starts, relevant context is returned ready to inject into your model's prompt
import { ConvoMem } from '@optraze/convomem-sdk'
const client = new ConvoMem({ apiKey: process.env.CONVOMEM_API_KEY! })
// Send conversation turns for background extraction
await client.capture(
[
{ role: 'user', content: 'I need help with my order #4471' },
{ role: 'assistant', content: 'I see your order was delayed. Let me fix that.' },
],
{ email: '[email protected]' },
{ channel: 'CHAT' },
)
// Recall relevant context before the next reply
const { context } = await client.lookup(
'order delivery status',
{ email: '[email protected]' },
{ limit: 10 },
)What's next
Since this post was first published, several items from our roadmap have shipped:
- On-prem deployment — available on request for organizations that need ConvoMem inside their own infrastructure; contact [email protected]
- Custom categories — organizations can define their own memory categories alongside the built-in ones
- More integration surfaces — the official TypeScript, Python, and Rust SDKs and a hosted MCP endpoint ship in v1.2.0
Stay tuned for updates.