About

A small agent. Never the one in charge.

It works the quiet hours and leaves a note. Anything your policy holds comes to you first, and every tool call is on the record.

12
bundles, one per trade
1
person per agent
0
models trained on your data
The name

About the cleaner wrasse.

Cleaner wrasses are reef fish that run cleaning stations. Larger fish — sharks, groupers, eels — voluntarily approach, queue up, and hold still while the wrasse picks parasites from their gills.

The relationship is so reliable that fish remember which station they visit and return to it. Marine biologists call this a mutualistic cleaning symbiosis. We call it a description of the product: small, trusted, useful enough to be sought out — and never the thing in charge.

Why we exist

The gap in personal AI is bigger than it looks.

Generic assistants are everywhere, and so are do-it-yourself agents — if you are technical enough to run your own server, vet every skill by hand, and have a weekend spare.

At the other end are enterprise platforms: powerful, priced per seat, and sold with a procurement cycle attached. Excellent for a company that has one. Most working professionals do not.

Wrasse is the middle: hosted so there is nothing to maintain, on the record so you can see what it did, and tuned for a trade so it knows your work before you explain it.

What we believe

Four principles.

Not a manifesto. The four things we keep coming back to when product decisions get hard.

Do the work

Don't sell the work. The product is what it does for you, not what it tells you about itself. Wrasse should disappear into your day, not announce itself in it.

Check before acting

Drafts are cheap. Sent messages are not. Every tool call is checked against your policy before it runs, and anything it holds asks you first.

Audit by default

Every tool call is logged and on the record. You can read what your agent did, export the log, and close your account to have your data erased, apart from what the Privacy Policy lists as kept or not yet erased.

Earn trust, then keep it

Trust is the moat. We work hard to earn it the first time, and harder not to break it. If we make a mistake, we say so. If you ask us to forget something, it's forgotten.

Where we're going

The next eighteen months.

Now

Closer for realtors, the first bundle. Building out the runtime, the audit log, skill curation, and the realtor integrations.

Next

The remaining bundles, each scoped with a design-partner cohort. Audit log v2, and a small-practice mode for two-to-five person firms.

Later

Public bundles from vetted third-party authors, a team SKU, and an on-prem option for regulated trades. Wrasse becomes a platform other people build vertical agents on.

The team

Who's building this.

The Wrasse team is small and getting bigger deliberately. The founding team has shipped infrastructure software, AI systems, and consumer products at scale. We're hiring engineers and a designer through 2026 — if that's you, say hello.

Investors and advisors are listed on request. We're privately backed, and we would rather talk about the product than the cap table.

Read how we handle security and your data →