Memory infrastructure for your AI agents
Mnexium is a memory management platform for LLMs. Store, score, and retrieve long‑term context with a simple API. Built for ChatGPT, Claude, and custom agents.
Built for ChatGPT, Claude, and custom agents
Long‑term memory in LLM apps is hard
Teams re‑implement vector stores, custom schemas, scoring logic, and retrieval for every app. Mnexium gives you a centralized, API‑first memory layer so you can focus on your product.
- Centralized memory store for all your agents
- Simple REST API and SDKs
- Built‑in scoring: importance, recency, visibility
- Governance: who can see what, and when
What you get
Memory layer
Store, Manage & retrieve memories with clean REST/JSON endpoints.
Dynamic Prompts
Compose context-aware prompts powered by saved memory and chat history.
Scoring & ranking
Importance & recency scoring, filtering, and search.
Chat governance
Attach memories, enforce visibility, and keep context safe.
How it works
Save History
With a few additional JSON fields, Mnexium provides a comprehensive foundation for AI agents—persistent memory, conversation history, relevance scoring, and governance controls—without additional SDKs or infrastructure to manage.
const client = new OpenAI({});
const response = await client.responses.create({
model: "gpt-4o-mini",
input: "Write a one-sentence bedtime story."
});
console.log(response);
const client = new OpenAI({})
const response = await client.responses.create({
model: "gpt-4o-mini",
input: "Write a one-sentence bedtime story.",
mnx: {
save: true
}
}
History API
Public endpoint to retrieve chat history by chat_id. Returns an ordered list of messages (oldest to newest). Supports an optional limit parameter.
https://api.www.mnexium.com//api/chat/history?chat_id=10b39141-4ff0&limit=200
{
"messages": [
{
"role": "user",
"message": "Write a one-sentence bedtime story.",
"message_index": 0,
"event_time": "2025-12-05 02:18:34.820"
},
{
"role": "assistant",
"message": "As the stars twinkled like tiny diamonds in the night sky, a little owl named Oliver discovered that if he spread his wings wide enough, he could soar through his dreams and dance with the moon.",
"message_index": 1,
"event_time": "2025-12-05 02:18:34.820"
}
]
}
Use History
Enable mnx.history to automatically prepend prior messages to your OpenAI chat call while saving the new turn. This keeps context persistent across interactions with zero extra plumbing.
const res = await client.chat.completions.create({
model: "gpt-4o-mini",
messages: [
{ role: "user", content: "Write another please." }
],
mnx: {
chat_id: "10b39141-4ff0",
save: true,
history: true
}
});
{
"messages": [
{
"role": "user",
"message": "Write a one-sentence bedtime story.",
...
},
{
"role": "assistant",
"message": "As the stars twinkled like tiny diamonds in the night sky, a little owl named Oliver discovered that if he spread his wings wide enough, he could soar through his dreams and dance with the moon.",
},
{
"role": "user",
"message": "Write another please."
},
{
"role": "assistant",
"message": "In a cozy corner of the forest, a gentle rabbit named Rosie found a magical garden where each flower whispered sweet secrets, wrapping her in warmth as she drifted off to sleep beneath the soft glow of the fireflies.",
}
]
}
Use cases
Personal AI assistants
Remember preferences, routines, and history across sessions.
Sales & support copilots
Recall tickets, CRM notes, outcomes — fast.
Developer tools & IDE agents
Persist codebase knowledge and user habits.
Productivity apps
Calendar context, notes, tasks, and recurring patterns.
Feature highlights
- Bring your own OpenAI API key
- API support: Chat Completions and Responses
- History Management
- Dedicated Playground page in sidebar
Built for developers
No infra to manage. Bring your own LLM — OpenAI, Anthropic, or self‑hosted.
// Simple: fetch memories for an access token
const res = await fetch('https://mnexium.com/api/memories/atk_xxx?limit=25');
const { data } = await res.json();
// feed into your LLM prompt ...
Use OpenAI SDKs with Mnexium by pointing the baseURL to our API and setting apiKey and project on the client.
import OpenAI from 'openai';
const openai = new OpenAI({
baseURL: 'https://mnexium.com/api/v1',
apiKey: process.env.OPENAI_API_KEY,
project: 'my-project'
});
const resp = await openai.responses.create({
model: 'gpt-4o-mini',
input: [
{ role: 'user', content: [{ type: 'input_text', text: 'Hello from Mnexium' }] }
],
mnx: { chat_id: 'demo', user_id: 'user_123', save: true, history: true }
});
console.log(resp.response?.output_text || resp.output_text);
import OpenAI from 'openai';
const openai = new OpenAI({
baseURL: 'https://mnexium.com/api/v1',
apiKey: process.env.OPENAI_API_KEY,
project: 'my-project'
});
const chat = await openai.chat.completions.create({
model: 'gpt-4o-mini',
messages: [
{ role: 'user', content: 'Summarize Mnexium features.' }
],
mnx: { chat_id: 'demo', user_id: 'user_123', save: true }
});
console.log(chat.choices?.[0]?.message?.content);
Roadmap
- Per‑org policies and chat governance
- Advanced ranking and RAG integrations
- Multi‑tenant memory and audit logs
- SDKs for JS/TS, Python, and more
Pricing
Early access
Start free while in beta. Generous limits for early adopters.
- • Fair usage limits
- • Access tokens & memory APIs
- • Upgrade paths for teams
FAQ
How do I use Mnexium with ChatGPT?
Create an access token, then call the token-gated /api/memories/[token] endpoint from your tool or GPT action to fetch context.
What is a memory?
A small JSON record describing a fact, preference, event, or note. Mnexium stores and scores memories for relevance.
How is Mnexium different from a vector DB?
Mnexium is an opinionated memory layer that includes scoring, visibility, and governance — not just embeddings.
Can I revoke an access token?
Yes. You can revoke tokens at any time; requests using revoked tokens will be denied.
Is Mnexium suitable for production apps?
Yes. It is designed for multi-agent, multi-user workloads and will offer enterprise controls as we progress.
What languages do you support?
Start with REST/JSON from any language. SDKs for JS/TS and Python are on the roadmap.