# Gravy: a local-first AI agent for personal finance

Canonical: https://mudpie.ai/companies/gravy/
Breadcrumb: [Home](https://mudpie.ai/) / [Companies](https://mudpie.ai/companies/) / [Gravy: a local-first AI agent for personal finance](https://mudpie.ai/companies/gravy/)
Author: Ali Abouelatta (https://mudpie.ai/authors/ali-abouelatta/)
Published: 2026-09-19
Updated: 2026-09-19
Research type: Company profile
Method: Company and accelerator sources checked 2026-09-19. Product claims are attributed to their sources; this is research, not a hands-on product trial.

## What it does

Gravy is a local-first AI workspace for personal finance. It connects banks, cards and investments, categorizes transactions, tracks subscriptions and bills, answers questions about spending, organizes tax records and can act on tasks such as transfers or cancellations. The [YC profile](https://www.ycombinator.com/companies/gravy) describes an agent that moves money, trades stocks and files taxes; the [homepage](https://gravy.finance/) presents the safer starting point as one local workspace for seeing accounts and understanding what changed.

The fit is a technically comfortable person with a fragmented financial life who wants an assistant that can act, not another read-only budgeting dashboard. That ambition is also the main constraint. Any feature that moves money, trades or files taxes needs explicit approvals, reversible actions, clear logs and a recovery path when a connection or categorization is wrong.

## Why I’d look closer

Gravy’s homepage says it runs locally on a Mac, with accounts, transactions and documents never leaving the device, and no cloud storage or remote servers. This is company-reported technical positioning, not an independent security review. It is worth examining because local-first behavior changes the trust and backup model: users may get stronger data locality, but they also need to understand updates, device loss and what happens when an action requires an external institution.

The founder background fits both sides of the problem. The [YC biographies](https://www.ycombinator.com/companies/gravy) describe Ali Tabba as a former KiranaKart co-founder whose company became Zepto, with experience at BlackRock’s iShares, and Max Church as a former lead AI engineer at Moonpig. That is consumer and financial-systems context, not a guarantee that the agent should be trusted with live capital.

## What I’d ask

Which institutions and account permissions are supported? Are transfers, trades and tax actions always proposed for approval, and can a user inspect the exact command before execution? I’d test a read-only connection first, then a sandbox or low-stakes action, and inspect local storage, export, backup, deletion and recovery behavior. The sources checked did not expose public pricing, so the cost of the full action surface belongs in the same conversation as risk.

## My editorial take

Gravy is interesting because it treats personal finance as an action system and makes local execution part of the product thesis. I would not recommend it from the promise alone. It earns a closer look if the local-first claim, permission model and approval boundary survive hands-on inspection.

## Quick facts

| Field | Sourced detail |
|---|---|
| Buyer fit | Technically comfortable consumers with multi-account finances |
| Product surface | Account aggregation, categorization, subscriptions, investments and taxes |
| Privacy claim | Company says data stays on the Mac and is not cloud-stored |
| Public pricing | Not exposed in the sources checked |

## Sources checked

Checked 2026-09-19.

| Source | Used for |
|---|---|
| [YC company profile](https://www.ycombinator.com/companies/gravy) | Product scope and founder backgrounds |
| [Gravy homepage](https://gravy.finance/) | Local-first claim and current product surface |

## Cohort context

Gravy is listed in Spring 2026. In our 2026-09-18 directory snapshot, 21 of 193 listed companies in that cohort have YC’s primary industry label Fintech (10.9%). This is a current-directory comparison, not an original intake count or a performance ranking. [Nine-cohort dataset](https://mudpie.ai/research/yc-cohorts-2026-09-19.json).

## Public website snapshot

Observed 2026-09-19T16:16:15.316Z in raw homepage HTML. This records visible metadata and advertised links, not agent execution or product quality.

| Signal | Homepage observation |
| --- | --- |
| Product description metadata | Observed |
| Canonical link | Not observed in this response |
| H1 or H2 heading | Observed |
| Typed structured data | Not observed in this response |
| Docs/developer link | Not observed in this response |
| Pricing link | Not observed in this response |
| llms.txt link | Not observed in this response |
| Markdown alternate | Not observed in this response |

[Public observations](https://mudpie.ai/research/yc-homepage-links-2026-09-19.json) · [Collection method](https://mudpie.ai/research/yc-homepage-methods/README.md). Missing links here do not establish that a capability or file is absent elsewhere.


## Author disclosure

I cofound Lazyweb and publish Mudpie. This is an owner-written publication, not an independent testing organization. Research notes distinguish observations, sourced reporting and editorial judgment.