Dashboard

Context engine and
self-learning memory
for AI agents.

Your agents build their own long-term memory, fully automatically. As they work, they add what they learn, update what changes and remove what is outdated, so it gets better over time. When they need something, our context engine understands what they mean and delivers just the right memories in milliseconds, without bloating their context window.

Read the docs

Copy the prompt and paste it into your agent. It sets up Memory for you.

import requests

api = "https://api.ragionex.com/v1/memory"
client = requests.Session()
client.headers["X-API-Key"] = "rgx_memory_..."

# Save any text, from one line to a long document.
saved = client.post(f"{api}/write", json={"content": """
I'm building a budget tracking app.
I chose FastAPI because I already know Python.
I'll use PostgreSQL for the database.
""".strip()}).json()

# Later, any agent asks; the context engine finds the memory.
status = client.get(f"{api}/status/{saved['id']}").json()
if status["status"] == "ready":
    found = client.post(f"{api}/search", json={
        "query": "Which technologies did I pick for my budget app?"
    }).json()
    print(found["results"])
else:
    print(status["status"])  # Processing or Failed.
const api = "https://api.ragionex.com/v1/memory";
const key = "rgx_memory_...";
const memory = async (path, body) => (await fetch(api + path, {
  method: body ? "POST" : "GET", body: body && JSON.stringify(body),
  headers: {"X-API-Key": key, "Content-Type": "application/json"},
})).json();

// Save any text, from one line to a long document.
const saved = await memory("/write", { content: `
I'm building a budget tracking app.
I chose FastAPI because I already know Python.
I'll use PostgreSQL for the database.
`.trim() });

// Later, any agent asks; the context engine finds the memory.
const { status } = await memory("/status/" + saved.id);
if (status === "ready") {
  const found = await memory("/search", {
    query: "Which technologies did I pick for my budget app?",
  });
  console.log(found.results);
} else console.log(status); // Processing or Failed.

Isolated where it should be.Shared where it helps.

Each project gets its own Vault, so its memories never mix with another's.Invite teammates and give every agent only the permissions it needs.

  • Your agent 1
  • Your agent 2

You decide who shares each project and what every agent can do.Click Read, Write, Edit or Delete to see how permissions work.

Personal Vault

128memories

Shared Vault

94memories
  • Your agent 3
  • Teammate A
  • Teammate B

It runs itself.And gets better every day.

With the Ragionex MCP server, your agents handle memory on their own.

  • Saves on its ownAgents store decisions, preferences and facts as they work.
  • Fixes what changedNew facts update the old ones instead of piling up.
  • Undo anythingEvery change and delete can be undone for 30 days.
  • Gets better with useEvery session adds to what all your agents know.
  • EuropeYour memories are stored in the EU, from the very first save.
  • AES-256Memory text is encrypted. A copy of the database alone can’t reveal it.
  • IsolatedEach project’s memory stays in its own Vault. Searches never cross into another.
  • ExportDownload everything you saved, anytime. You are never locked in.

Start free. No credit card.10,000 requests a month.

Every save, edit or search is 1 request.Browsing, viewing, exporting and deleting your memories are free.

Read the docs