Axiom shipped version 5.1.0 on July 20, 2026, and what caught our eye was the new 2FA steps, because they show how the product thinks: a bot built from steps needs a step for everything, including the moment a site texts you a code.
That’s Axiom.ai in a sentence: a no-code builder where you assemble a bot from steps (open this page, click that element, copy the table into a sheet), plus LLM steps when you want a model in the middle. Around 100k people use it, it holds a 4.5-star rating, and we get the appeal: the bot does exactly what you told it, in the order you told it.
dassi goes the other way round. You describe the job in plain English, or do it once while dassi follows along, and it saves a workflow you can run on demand or on a schedule. There is no step list at all. Some people find that a relief. Others miss seeing every click laid out in advance, which is fair.
Redesigns are where step bots hurt
A step-based bot is pinned to the elements it was built against, so when a site moves things around it tends to break, and someone has to open the builder, find the failed step and re-point it. Axiom is sold as no-code. Still, a recurring complaint from people who leave is that repairs take some feel for the DOM and selectors. Not much, but enough to sting on a Monday when the weekly export didn’t run and whoever built the bot is on holiday.
dassi keeps no selectors. It reads the page on every run and works from what’s in front of it, roughly the way you’d cope the morning after a redesign (the Export button moved into a menu, so you open the menu). It can still get stuck. When the job itself changes, a new approval screen or a form that suddenly wants a cost center, dassi stops and asks you instead of guessing, and a model making choices on each run is less predictable than a fixed list of clicks. We think that’s worth it for messy work, and we’d probably still build a very rigid, high-volume job in Axiom, though we’ve never found a clean rule for where one kind of job ends and the other begins.
Logging in, locally and in the cloud
Axiom’s desktop runner uses the Chrome you’re already logged into, which is the easy case. Its cloud runs need your session cookies stored with Axiom, and cookies expire, so somebody resyncs them by hand, or you script the login as steps (hello again, 2FA steps).
dassi sidesteps that by staying in your own browser, Chrome, Edge or Brave, on the accounts you’re already signed into. If a session has lapsed it signs in with whatever Chrome’s password manager autofills, so the password stays in the password manager and you never paste it into a chat or into dassi’s settings. 2FA codes are yours to type, and dassi waits however long it takes you to find your phone.
When to choose Axiom
- You need runs while your laptop is shut, which Axiom’s cloud handles; dassi’s scheduled runs happen in your browser, so it has to be open.
- The job never varies and you want identical clicks in identical order.
- You like building step by step and don’t mind the occasional repair.
Pricing is by runtime: 2 free hours, then plans from $15 for 5 hours up to $250 for 250 hours. Bots that finish in seconds stretch that a long way; for long daily jobs we’d do the arithmetic first.
When to choose dassi
We built dassi for jobs that are miserable to express as steps and easy to explain to a colleague: pulling figures from three dashboards into a spreadsheet, or filling the same portal form with different data every week. Free credits to try dassi, good for 7 days. No card required.
You pick the model (Claude, GPT, Gemini, Grok, DeepSeek, Kimi, anything on OpenRouter, or a local model) and either bring your own key for $10 a month or take a plan with credits included, $25 to $300 a month. The agent runs locally. The instructions and page content it needs go to the provider you chose, through dassi’s relay when you’re on included credits, and anything you’ve flagged, like submitting or sending, waits for your OK.
If your Axiom bots have gone a whole year without a repair, leave them alone. We’d like to know which sites those are; we haven’t met many that sit still for a year.