Bloome Reviews 

The accounting/credits is in a total mess!!! All seemed fine until my initial credits ran out ("exhausted" was the word used). So I bought more and my bank account was debited. No receipt, no nothing. I was still stuck in the "exhausted" mode for two days even though I apparently now had a credit balance. Then I got some more daily credits, so complained to the Assistant, who pinged a total of 13 times and NO response - but I used up the daily allowance complaining! I have received no daily credits for two days now and apparently it has been set to NULL. Emails to hello at bloome have had no reply. I have a half-built website that I would like to deploy but lost all trust in bloome

Bloome is a solid product. The multi-model routing works well, the interface is straightforward, and the agent coordination is more advanced than most in this category. I've used it quite a bit, and I hoped to like it more than I do.

I started checking what the web client was doing because it kept crashing. The assistant would go offline mid-session or stop working entirely. What I found while I was in there went well beyond the crashes.

The first concern is privacy. The authenticated chat page loads a full advertising and social tracking stack. Facebook, Google Ads and remarketing, Google Analytics, TikTok, Twitter. The Analytics pageview includes the conversation ID, so conversation identifiers are going to Google. That's unremarkable on most products. In a private messaging tool, it's a category error, and under GDPR and CPRA, it isn't a small one. The app also lacks any security response headers, which is an edge configuration fix. Beyond that, I found issues in how credentials are handled that I reported privately and won't detail here.

The delete account link doesn't work. You click it, nothing happens. On a product handling private conversations, a broken erasure control is worse than not offering one.

The product is passwordless, which is defensible, but it means the session token is the only credential and there's no way for a user to end a session. If a token is exposed, the user has no lever at all.

What actually moved my rating wasn't any single issue. I filed 21 items over five days, several of which were marked urgent. Two got replies, both frontend. Security items sat in the same queue as cosmetic ones, with no separate path, no acknowledgment, and no stated response time. That isn't a bug. It's a missing process that predicts how a team will handle the next thing.

None of this is surprising for a product at this stage. What matters is whether someone is on the other end when a user reports something. Right now it doesn't look like there is.

I'll revise this upward the moment that changes. The product deserves it.