Back to articles
How to set up cross-agent memory for your finances
You told Claude about your savings goal last week. Now you open ChatGPT and have to explain it all over again. That's the problem cross-agent memory solves — and setting it up takes about five minutes.
This guide walks you through connecting Era Context to your AI of choice so your financial context follows you everywhere, automatically.
What you'll need
- An Era account (free to start — no credit card required)
- A bank account at a supported institution
- An MCP-compatible AI client: Claude, ChatGPT, OpenClaw, Cursor, or Gemini all work
If you already use Claude for financial questions and want to understand what memory actually stores, what Era Context stores and how to manage it covers that in depth. If you just want to get set up now, keep reading.
Step 1: Create your Era account and connect a bank
Go to era.app and sign up. The Basic plan is free and includes two connected accounts and up to 50 stored memory facts — enough to get a feel for everything before you commit to anything.
Once you're in Era Context, tap the connect flow and link a bank account. Era connects through MX, a regulated financial data provider, so your bank credentials are never stored by Era. You authenticate directly with your bank, complete any two-factor step your bank requires, and your accounts appear in Era Context within about 30 seconds.
If you have checking and savings at the same bank, both show up automatically.
Step 2: Add Era Context to your AI client
Every MCP-compatible client follows the same basic pattern: add Era Context's MCP server URL and complete an authorization step. Here's how it looks for the most common clients.
Claude Desktop
Open Claude Desktop, go to Settings → MCP Servers, and add a new server:
{
"mcpServers": {
"era-context": {
"url": "https://context.era.app"
}
}
}
Save, and Claude will prompt you to authorize the connection. Once you confirm, Claude has live access to your financial data and your memory profile.
ChatGPT
In ChatGPT's settings, look for the MCP or external connections section and add https://context.era.app as the server URL. Complete the OAuth consent screen that appears.
Other MCP-compatible clients
The URL is always https://context.era.app. Any client that supports MCP follows the same pattern: paste the URL, complete the authorization, and the connection is live.
Step 3: Start building your memory profile
Here's the part that surprises most people: you don't need to set up memory explicitly. You just talk to your AI like you normally would, and Era stores what matters.
Try it. Open a conversation with your connected AI and mention something real:
- "I'm saving $800 a month toward a house down payment — I want to be ready by next spring."
- "My gym membership is essential spending, not discretionary — please don't flag it as something to cut."
- "My partner covers the mortgage, I cover utilities and groceries."
Each of these gets stored in your Era Context memory profile. Now open a different AI client and ask about your finances — it already knows, without you repeating yourself.
On the Basic plan, your memory profile holds up to 50 facts — enough for a solid financial context across goals, preferences, and household arrangements. The Organize plan expands that to 200 facts, and Automate and above remove the limit entirely. See pricing if you're curious about what's included at each tier.
Step 4: Check what's stored on your Knowledge page
Era's Knowledge page (at /app/knowledge) shows everything currently in your memory profile: your stated goals, saved preferences, facts your AI has learned from your conversations, and any inferences it's made that you've confirmed.
From there you can:
- Review what's stored so far
- Edit any fact that's outdated or wrong
- Retract anything you want removed — it disappears across all connected agents instantly
You're always in control of what your AI remembers. If something in the list no longer applies — say, a savings goal you've already hit — you can remove it with one tap, and it's gone everywhere.
Step 5: Connect additional AI clients
One Era account works with as many AI clients as you want. Connect Claude for in-depth analysis, ChatGPT for quick questions, Cursor for financial scripting while you work — each one reads from and writes to the same memory profile.
When you connect a new client, it inherits your full existing memory profile immediately. No setup, no re-explaining your situation from scratch. The new assistant already knows your goals, your preferences, and your context.
To connect another client, repeat Step 2. Each connection requires its own authorization step — you control which clients have access, independently.
What cross-agent memory actually feels like
The moment that makes it click is usually the first time you switch clients mid-task. You start a conversation with Claude about whether you can afford a vacation without derailing your savings goal. Claude knows your goal because you mentioned it two weeks ago. You switch to ChatGPT to run a quick scenario. It also knows your goal — same Era Context, different client.
For a deeper look at why this portability matters and how it compares to what built-in AI assistants offer, why your financial AI memory should be portable walks through the tradeoffs clearly. And if you want to understand the full picture of what gets stored and how sharing works across agents, can AI agents share financial memory? covers the mechanics.
FAQs
How long does setup take? About five minutes from account creation to having an AI client with live bank access and memory working. Longer if your bank has a complicated two-factor flow, but rarely more than ten minutes.
Does memory work on the free plan? Yes. The Basic plan includes up to 50 stored memory facts with full read and write access — you can review, edit, and retract any stored fact. Paid plans add higher limits: Organize gives you 200 facts, and Automate and above remove the cap entirely.
What if I want to start fresh? Head to the Knowledge page and retract whatever you want cleared. There's no "reset everything" button (to protect against accidents), but removing individual facts is fast, and each removal is instant across all connected agents.
Which AI clients work with Era? Any MCP-compatible client — including Claude, ChatGPT, OpenClaw, Cursor, Manus, and Gemini. When a new client adds MCP support, it works with Era immediately.
Is my memory private? Your memory profile is private to you. It's never shared with other users, never used to train AI models, and only accessible to the AI clients you've explicitly authorized. You can revoke any client's access at any time from Era Context settings.