Rebel is a desktop AI workspace for agentic work. It connects your memory, meetings, files, actions, automations, and tools so AI agents can help with real work — while keeping sensitive actions behind approval checks. Built Fair Source, with portable workflows and model choice.
Since it's local-first with model choice, can the approval checks for sensitive actions run entirely on a local model, or does the gating logic still require a cloud model call?
Mindstone Rebel
@nikita_zaro You are two steps ahead! That is exactly the kind of thing we're thinking.
Right now, there are 3 different tiers of model:
- Planner for overall task decomposition, needs a smarter model
- Worker does most of the heavy lifting, could be a mid-tier model
- Background/behind-the-scenes model: does a lot of the lower-level house-keeping (e.g. safety-approval classification).
You could set that background/BTS to a local model, and then the safety approval decisions would be routed through it.
I haven't tried this, but it's a great idea.
@greg_detre Love this — the three-tier setup is really elegant, and routing safety approvals through a local background model makes a lot of sense. Thanks for the detailed reply!
Mindstone
@nikita_zaro you can configure it to entirely rely on your local model even for the gating logic
@joshuawohle That's great to hear — full local model support even for the gating logic is exactly what I was hoping for. Thanks!
The 'asks first' framing is smart — most AI agents just barge ahead and you find out later it did the wrong thing. Is this confirmation step configurable, or always-on by default?
Mindstone
@martin_mo it's on by default, but it learns from what you approve and you can tell it things like "always approve actions like this in the context of Y".
@joshuawohle That's a smart middle ground — letting it earn autonomy gradually instead of going full-auto from day one feels a lot safer for trust-building. Are most users defining those 'always approve X in context of Y' rules upfront, or do they tend to add them reactively after seeing what it gets right/wrong a few times?"
The ask-first posture is important. For small teams, the useful split is usually reversible vs irreversible actions: summarize, draft, search, and organize can move fast; writes to customers, CRM, payments, deploys, or files that other systems trust need a visible checkpoint and a trace.
Mindstone
@krekeltronics totally agree. Rebel also allows you to switch certain tools off entirely if you’re uncomfortable with it using them (like sending email vs drafting email)
Mindstone Rebel
@krekeltronics Yes, exactly.
You can also take a crude approach, e.g. "only enable read-only tools for Gmail", or something more sophisticated (see other comments).
Our built-in MCPs know which tools are read-only vs read-write to make this easier.
And if you're the technical admin for a Rebel rollout to your company, you can disable certain tools for everyone.
Nice approach. How's conflicting context handled when agents pull from multiple different projects?
Mindstone
@dhiraj_patel5 great question: it tries to triangulate, but ultimately will surface to the user if it cannot decide what is correct.
Personally, I have a daily automation running to clear up discrepancies between data sources and ask if unsure.
"For all of it, not one task" is the right framing, most AI tools
optimize for single workflows and break the moment your work spans
multiple systems.
How does Rebel decide what needs approval vs what can run autonomously?
The friction between trust and speed is where most agentic tools either
slow you down or scare you.
Mindstone Rebel
@elias_motionfy Hi Elias, it’s definitely a point of friction. Rebel starts off cautiously by asking for approvals quite a bit, but it learns your preferences from each approval or rejection. Over time, this means it can operate more autonomously and ask for fewer approvals.
"Ask first" is probably the right default. The challenge is that the more context an agent has, the more confident people become in its actions. Have you found a point where users start approving things without really reviewing them? Curious how you're thinking about trust calibration over time.
Mindstone Rebel
@jared_salois Hi Jared. We’ve definitely seen users build trust in Rebel over time. The goal of our safety system is for Rebel to learn what you do and don’t want it handling autonomously, so it only interrupts you when it genuinely needs input.
The planner/worker/background-safety-classifier split is a genuinely interesting routing decision. When the safety classifier and planner disagree on risk - does planner intent override a conservative flag, or does the classifier always win?
Mindstone Rebel
@taylor_time Hi Kevin, Rebel is designed to never overwrite the safety controls. If the planner wants to do something that your safety settings prohibit, it has to stop and ask you. If you decide not to approve the action, then the planner will work with you on what to do instead.
@liam_collins2 That's nice and clean. Does the planner surface why it flagged something, or is it a silent hard gate?
Mindstone Rebel
@taylor_time Yes absolutely. The user sees an approval notification explaining what the agent was trying to do and why it felt it needed to ask for approval.