Mindstone Rebel

The ask-first default is strong. The piece I’d want durable is the rule change after an approval: original action, future scope, who accepted it, and when it should expire or be reviewed. Otherwise learned permissions can quietly become permanent permissions. Are those rule updates inspectable later?

The approval-before-action model works well for individual workflows but gets complicated fast in team settings. Imagine two people on the same team have given Rebel conflicting learned rules for the same action type. One has approved it, one has flagged it. When a shared automation runs, whose learned policy governs? This seems like it could silently diverge across team members over time, especially as each person's rule set adapts independently. Have you thought about a policy reconciliation layer for shared workflows, or is the current model intentionally keeping rules personal and non-shared?

@melissanthi_papacharalampous1 The “ask first” layer is the strongest part for me. Agents become much more useful when they can work freely on low-risk tasks, but pause before actions that touch email, shared spaces, or anything sensitive. That middle ground between locked-down and fully autonomous feels like where real trust starts.

If I’m understanding this right, the local-first/customisable thing makes sense, especially if the point is not getting stuck inside one AI vendor’s way of working.One thing I’m wondering though is what happens once this moves from one person using it to a small team using it. If people start adapting their own workflows, connectors and approval rules, do you risk everyone ending up with slightly different ways of working?Is there a way to share the setups that work well without making it feel like one central locked-down workspace again? Curious how you’re thinking about that.

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?

What kinds of actions can Rebel actually execute today, like email and calendar, or internal app automations?

"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.

"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.

Congrats to the whole Mindstone team, this is a serious piece of work! Really like the approval layer being impact-aware, writing to a shared space treated as riskier than a private one is the kind of detail most agent tools never bother with. We've been building an agent that takes real actions and know the gate is the hardest thing to get right, too loose and its dangerous, too tight and nobody uses it. Rebel learns rules from approvals instead of re-asking forever and you clearly ran this on a lot of real work - very impressive. Another great thing is tool-level permissions like an email MCP that can draft but not send - IMO exactly the granularity people actually need. Nicely done!

Strong angle, and the maker replies show you've thought about the hard part. One thing I'd add from doing this on customer-facing work: the risk with ask-first isn't only where you draw the line, it's approval fatigue. If the agent interrupts too often, people start rubber-stamping every prompt and the gate quietly stops protecting anything. Two things seem to matter most: making each approval information-rich enough to decide in a couple of seconds (what it's about to do, why, and the source it's acting on), and the auto-updating rules you mentioned so it stops re-asking about things someone has already approved a dozen times. Get those right and ask-first scales; get them wrong and it becomes click-through noise. Curious whether Rebel surfaces the "why" and the source inline at the approval moment. Congrats on the launch.

the approval checks for sensitive actions is something more agent tools should be doing. most platforms either give the agent full access or keep it so restricted it can't do anything useful. that middle ground where the agent works freely on low risk stuff but pauses for human approval on anything sensitive is how you actually get teams to trust agents with real workflows. curious how granular the approval rules are, can you set different thresholds per agent or per action type?

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.

The 'ask first' principle is something we've had real debates about internally. Autonomous agents that act without confirmation can compound errors in ways that are hard to recover from, especially in stateful workflows. What's your approach to calibrating when the agent should interrupt vs. proceed? Does it use confidence thresholds, or is it more rule-based?

Nice approach. How's conflicting context handled when agents pull from multiple different projects?

Love the Fair Source approach. Curious why did you choose fair source over fully open source for Rebel, and what feedback have you received so far from developers?