Axiom.ai Remembers Your Cookies. It Does Not Remember Your Bot.

Axiom.ai is one of the most approachable browser automation tools out there. You build the bot visually with a Chrome extension, clicking through the steps the way a person would, and it runs on a schedule in the cloud: every morning it opens the pages, fills the forms, scrapes the tables, and drops the results somewhere useful, usually a Google Sheet.

So when tomorrow's run starts, how much does it remember about today's?

The honest answer: it remembers your login. It does not remember anything about what happened. That gap matters once your bots run on autopilot for months, because "the bot ran fine" and "the bot knew what it was doing" turn out to be two very different things.

What Axiom.ai actually persists

Start with what is genuinely there, because the platform is unusually explicit about its data posture.

Stored cookies and logged-in sessions. In the run options panel, you can choose to store cookies and share browser cookies and local storage with cloud runs, so a scheduled run reuses your logged-in session instead of logging in from scratch each time. That is the memory most users notice: your bot does not get stuck at the login wall on run two hundred.

The run schedule and configuration. You set a start date, time, and frequency, plus timezone, proxies, max runtime, headless mode, concurrency, and notifications for failures or completions. All of that is configured once and reused forever.

Google Sheets as the unofficial state store. Axiom.ai bots are built around data sources: a Google Sheet, a webhook, an API call. The typical pattern is read rows in, do the browser work, write results out. Operators end up keeping their bot's "memory" as rows in a sheet: processed IDs, last checked timestamps, error flags.

Local desktop runs. If you run on the desktop app, nothing leaves your machine. That is a privacy property, not a memory feature, but it matters for where state can live.

What starts from zero every run

Here is the other side, and Axiom.ai documents it plainly.

Every cloud run is an isolated session that is never reused. That is the platform's own phrasing. The run executes in a private container, and when the run finishes, the container is destroyed. Nothing about the working state of that run survives: no notes about what the bot saw, no record of a page layout change it worked around, no trace of a decision it made.

Data passed into a run is not stored. Their FAQ says it outright: axiom.ai does not store user data, and data passed in from a data source, a webhook, or an API call is not stored. So the sheet rows your bot read on the way in? Not retained. The webhook payload that triggered it? Gone. The platform processes and forgets, by design.

Run-to-run context is your job, not the platform's. There is no cross-run working memory, no "here is what happened last time, adjust accordingly." Each scheduled run wakes up with its configuration and its cookies, and nothing else.

This is not a flaw in the product. It is a privacy-forward architecture choice, and for simple deterministic bots it is exactly right. But it leaves a specific hole once the work gets even slightly judgmental.

The hole: your bot re-learns the same lessons on a loop

Picture a bot that monitors supplier pricing and flags meaningful changes. Monday's run encounters a supplier page that changed its layout; the bot's steps still work, but it takes a different path to the price, or a human edits the bot to handle the new layout. Tuesday's run has no idea any of that happened. If the layout flips back on Thursday, Friday's run re-derives the workaround from scratch, and if it fails, you read the error cold, with no record of what was tried before.

Or take error triage. A bot fails at 3 AM with a timeout on step seven. Was that the same timeout it hit last week, the one you decided was transient and safe to ignore? There is no memory of the decision, so you triage it from zero again, at 3 AM.

Google Sheets patches some of this. You can write a "notes" column, a status column, a retry counter. Plenty of operators do exactly that, and it works, up to a point. But a sheet row holds facts, not reasoning. It does not hold why you decided that timeout was safe to ignore, or what the page layout looked like when the bot adapted, or which supplier pages are chronically flaky and deserve a looser threshold. And every new bot you build needs the same plumbing rebuilt, in the same brittle way.

The pattern that closes the hole

The durable fix is a memory layer that sits outside every run, in a place every run can reach:

1. Each run starts by asking what is known. Before the bot acts, it pulls the relevant state: what happened last run, what was decided, what is known to be flaky or safe to ignore.
2. Each run ends by writing back what changed. Not a log line. The findings, the decisions, the reasoning, in a form the next run can actually use.
3. The memory is shared, not per-bot. When your pricing bot learns that a supplier's layout flips weekly, your stock-check bot should know that too, without you copying sheet formulas between projects.

This is the gap Vilix AI is built to fill. It is a cloud-hosted memory layer, so there is no database to maintain and nothing to install alongside your bots: every run reads from it and writes to it. Because it connects over MCP, the same memory follows you from Axiom.ai to your other tools, your coding agents, your chat apps. One memory, every tool.

And it stores full conversation history, not just the facts somebody decided were worth keeping. That distinction matters for exactly the 3 AM scenario above: you can go back to the actual exchange and re-derive the reasoning, not just the conclusion. Your data stays yours: export everything or delete it anytime in a portable format. The free plan is free forever, and the Pro trial runs seven days with no credit card.

Axiom.ai will keep remembering your cookies, which is what it is for. For everything your bots learn in between runs, put the memory somewhere the next run can actually reach.

Sources: Axiom.ai run options and scheduling (docs), data handling FAQ (docs), cloud and desktop run architecture (infrastructure).

Learn how Vilix AI gives your scheduled automations shared memory