You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Grok Bot and Alternatives chapter: a hosted, non-gated option with explicit approval gates (Krafty Bot) #287
Thanks for the Grok Bot and Alternatives chapter. It is the only survey of this category I have found that works from primary sources and says what it could not verify.
I build Krafty Bot (https://kraftybot.app), and I think it belongs in the chapter's comparison, with the caveat below.
Shape. Same as Grok Bot: named Bots with their own memory and working instructions, one persistent cloud computer per account shared by all of them (desktop and browser, files and browser sessions carry over), routines on a schedule or on an event in a connected app, group chats and handoffs. Connected apps are used through the owner's own sign-ins.
Where it differs from Grok Bot.
No plan gate. Free plan (one Bot, one connected app, a one-time $1 AI allowance, 20 minutes of computer time) and a $20/month Pro plan (up to 10 Bots, unlimited apps, $10 AI allowance, 50 hours of computer time). Runs in the browser on any OS; no desktop app required.
Approval boundary is explicit and documented: every write to a connected app waits for approval unless the owner always-allows that action for that Bot; sending and posting always ask; high-risk actions can never be always-allowed; on the computer, a separate vision model checks each click and keypress against a screenshot before it runs; nothing in a web page or email can approve an action; requests expire after seven days. Source: https://kraftybot.app/ai-agent-human-approval
Spend is visible per reply, with per-task limits.
The caveat for your lab. It is hosted and closed source. There is no self-hosting path and no choice of where the computer runs, so against the "local or lab-hosted compute, model choice" priorities in section 4 it scores like Grok Bot, not like Rakazo. If that disqualifies it from the chapter, a one-line mention in the hosted group with that limitation stated is what I would ask for.
Product facts in plain text, kept in sync with the code that enforces the limits: https://kraftybot.app/llms.txt. I am happy to answer questions or set up a Pro account if you want to verify any claim. Not affiliated with xAI.
Activity