Hi HN,
I built Era after getting frustrated that every fintech app wanted to be its own AI chatbot.
I already use Claude for everything. I didn't want to learn another app's interface. I wanted to ask Claude "did any subscriptions increase this month" and have it actually know.
Era is an MCP server that connects to your bank accounts through Plaid and exposes read-write financial tools to any MCP-compatible AI. You can check balances, send transfers, set automation rules, and manage budgets through whatever assistant you already use.
One thing we spent real time on: granular permission scopes. Read access, transfer limits, rule-setting — each is a separate capability so you can give Claude read-only while letting Agency, our proactive monitoring app, handle automated actions.
Thesis is a separate product we built on top for self-directed investors who want quantitative portfolio analysis.
Still a lot to figure out on the trust and safety side — would genuinely love to hear how people here think about AI write access to financial accounts.
IMO this is brilliant, necessary, and the short answer to your final question there is 'not in any world I can imagine right now.'
There are simply no 100% reliable guardrails or models or combination thereof, and without 100%, I wouldn't use it for the risk factor, and I wouldn't bet on it for the liability factor.
But read-only, able to warn me about upcoming drafts, note recurring subscriptions that seem stale, remind me which bills are coming due without me asking?
Yep. All day. Take my money.
As a person on the spectrum, I need this badly enough that I've been working on the same for myself with no intentions of marketing it.
I'm looking forward to trying yours, and maybe saving myself the effort! Cheers!