Memra gives your AI a permanent memory — whether you're building AI apps with our API or using AI tools in VS Code with our extension.
Two products, one platform
Memory API
For AI apps you build
VS Code Extension
Auto-captures AI chats
Three steps. Two minutes. Production ready.
Add the npm package to your project
npm install @memra-client/clientStore every exchange with one line
await memory.save(userId, userMsg, aiReply)Semantic search finds relevant memories
const { context } = await memory.getContext(userId, query)User frustrated. Context lost.
AI remembers. User delighted.
Battle-tested infrastructure for production AI memory.
~125ms average. Semantic search powered by pgvector. Fast enough for real-time conversations.
Not just last N messages. We find the most RELEVANT memories using vector similarity, not recency.
Works with GPT-4, Claude, Gemini, Groq, Llama. One SDK. Any model.
Separate memory spaces per agent ID. Customer support bot ≠ coding assistant.
API key auth. Data isolated per key owner. GDPR-compliant deletion built in.
Two lines of code. No infrastructure to manage. Just install and go.
Whether you're building AI apps or using AI in VS Code — Memra gives your AI a memory.
Add persistent memory to any AI application. Save conversations, retrieve context, and give every user their own memory layer.
Auto-captures every AI conversation in VS Code. Works with Copilot, Claude Code, Cline, and Continue. Zero setup — install and forget.
Not sure which? Building an AI app → Memory API · Using AI in VS Code → Extension
Install once. The extension handles everything.
One-click install from the VS Code Marketplace. Paste your Memra extension key and you're done.
Use GitHub Copilot, Claude Code, Cline, Continue — the extension works silently in the background.
Every conversation is saved. When you start a new session, your AI picks up exactly where you left off.
Works with
GitHub Copilot · Claude Code · Cline · Continue · and more
Without Memra
Re-explain your project every session — 10-20 minutes wasted
With Memra
AI picks up exactly where you left off — instantly
Drop in one function call. Remove the rest.
import { MemoryClient } from '@memra-client/client'
const memory = new MemoryClient({ apiKey: process.env.MEMRA_KEY })
export async function chat(userId: string, userMessage: string) {
// Get relevant past context
const { context } = await memory.getContext(userId, userMessage)
// Memory goes in system prompt — NOT in the user message
// This tells the AI what the context is and how to use it
const systemPrompt = context.length > 0
? `You are a helpful assistant with memory of past conversations.
Here is what you remember:
${context.map((m, i) => `${i + 1}. [${m.role}]: ${m.content}`).join('\n')}
Use this memory to give personalized responses.`
: `You are a helpful assistant.`
// Your AI call — works with any provider
const reply = await openai.chat.completions.create({
model: 'gpt-4o',
messages: [
{ role: 'system', content: systemPrompt },
{ role: 'user', content: userMessage },
],
})
// Save to memory
await memory.save(userId, userMessage, reply.choices[0].message.content ?? '')
return reply
}Start free. Scale when you're ready.
Perfect for side projects
For production applications
For teams and high volume
Be the first to know when it launches
memories stored
and counting
avg latency
semantic search
uptime
SLA guaranteed
to integrate
seriously, that's it
Join developers building smarter AI applications with Memra.
Get started for freeNo credit card required · Free forever plan