Launching today
qorqbot
Proactive agent for business work
96 followers
Proactive agent for business work
96 followers
qorqbot is proactive qorqbot runs on subscriptions you already pay for - Claude, Codex, and more. It is provider-agnostic and built for business workflows and complex, long-running tasks qorqbot manages your attention, and gets work done in the background. it learns from you, without you actively teaching it, and synthesizes your preferences, workflows, and context naturally qorqbot is not a harness - it sits around your harness, providing operational, context, UX, and self-adjustment layers
Launch tags:Developer Tools
Launch Team / Built With
AWS Activate CreditsCredits that fuel your startup’s next milestone
Promoted
Hey PH!!
Its Alp again - presenting Qorqbot.
Qorqbot is a provider-agnostic proactive agent for business workflows and complex, long-running tasks, powered by the AI subscriptions you already pay for (codex, claude, etc.).
It learns how your processes, hierarchy, context, and the way you like things done.
It carries work forward in the background and brings you the decisions that need your attention.
Signoffs gate sensitive actions and decisions, pending your approval.
Its self-learning system adapts to your needs, preferences, and culture as you work with Qorqbot.
Operational layer provides UX for you and the agent to work with clients, CRMs, tables, and communications.
Agentic layer gives you natural way to build custom skills, workflows, and tools with easy plug-and-play architecture.
Qorqbot started as a personal assistant for me, and main issue with top labs' agents was the provider lock-in. I do not want to build operations for my business on Grok, just for them to hike the rates. I also am not a fan of sharing my very personal data.
I designed the system around these features:
Proactive Agent:
I want my guy to only stop working when there is not much else left to do. I want him to reply to emails as they come, manage my clients relationships (we have integrated CRMs), and be able to execute multi-step protocols/workflows.
Easy integrability:
Since the code is open, anyone can adjust the qorqbot for their usecase at any depth level - they can add a new layer, or add a new operational object type, or adapt the chat interface to use their custom models.
Deep Tailoring & Protocolization:
Overseer (special agent meant only for tailoring the system to you) agent reads your transcripts, threads, and comments, and suggest "deltas" to the agentic layer. This goes beyond preferences - Overseer has full ability to add new skills, make new plugins, formalize workflows, create and edit tools. It has full control over the harness that Qorqbot runs in.
Its still early access, and I run full operations entirely myself (+ qorqbot), and I am actively looking for bright partners!
Overseer is the part I'd want. I do it by hand weekly, reading my agent sessions and turning misses into rules. does it ever drop a rule that went stale?
@olia_nemirovski Thanks, Olia! Your weekly session review is exactly the kind of work I want Overseer to help with: noticing repeated misses and turning those lessons into something the agent can use next time. Your question about stale rules matters just as much. A useful rule can become wrong when a client, workflow, or preference changes.
My view is that rules should be revisable, and the system should preserve enough context to explain why a rule exists and what would justify changing it. I’d be cautious about removing something simply because it hasn’t been used recently, especially if it represents a permission boundary. A conflict with newer instructions is a much stronger reason to surface it for review. That’s the standard I’d want for this kind of adaptation: useful learning over time, with clear control over what gets retained.
Curious how you handle that at Tobira today—do you revisit old rules in your weekly review, or mostly notice them once they start causing problems?
running on the claude and codex subscriptions people already pay for is a smart call. congrats Latif!
@mohitvishwakarma Thank you, Mohit! That was a big part of the idea behind Qorqbot. If someone already pays for capable models through Claude or Codex, I want them to be able to put that capacity to work on the operational tasks that keep piling up throughout the day.
What interests me most is everything around the model: keeping the relevant context, carrying a task forward, learning from corrections, and knowing when to bring a decision back to you. There’s a lot of value in making those pieces work together for the things people actually need done. Being able to start with a subscription they already use makes it easier to try Qorqbot on a real workflow and decide whether it earns a place in their day.
We’re still early, so I’m especially interested in what people choose to hand over first and where they want more control. Really appreciate you checking it out and supporting the launch!
The signoffs are what would make or break this for me. I run everything solo, so the boring middle — triaging, drafting, chasing — is exactly what I'd want to hand off, but only if I can trust the gate on anything client-facing. Curious how it decides what needs a signoff versus what it just does, and whether the Overseer learning from my transcripts ever picks up a habit I'd rather it didn't.
signoffs gating the sensitive actions is what would make me ok letting it answer client emails, nobody wants an agent replying fully on its own in week one
when Overseer suggests a delta to the agentic layer, do you review it before it goes live or does it apply right away??? feels like the place where one small mistake spreads acros every workflow
@thamibenjelloun Thanks, Thami! I agree that client email is a workflow where people need to build confidence gradually. Having the agent prepare a response, then bringing back the actual draft for a decision, is a useful place to start.
On the Overseer question, I’d treat a change to shared agent behavior as something that deserves its own review. The review should make clear what is changing, why the change was suggested, and which workflows it could affect. Anything that alters permissions, approval requirements, or how the agent communicates with clients should need explicit approval before it takes effect. A suggestion based on one interaction needs particular care if it is going to influence work across multiple clients.
I’d also want changes to be easy to inspect and undo if they produce the wrong behavior. For me, the point of adaptation is to reduce the amount of supervision you have to do over time, while keeping you in control of the decisions that carry consequences. Thanks for pressing on that distinction!
Bringing decisions back for a signoff is the part I'd pay attention to. Keeping the approval tied to the exact draft or action would help when background work changes between the request and the decision.
@oleksii_sekundant Thank you, Oleksii! That detail matters a lot. If I approve a client email, the approval should refer to that specific message, addressed to that specific person, with those attachments. If the recipient, attachments, or substance of the message changes while background work continues, that should trigger a fresh review before anything gets sent. Otherwise, the original signoff stops representing the decision I actually made.
I also think the approval request itself should make those details easy to see, including what changed since a previous draft. People need enough context to make an informed decision without having to reconstruct the agent’s entire session.
That’s the level of control I want people to have as they give Qorqbot more responsibility: let research and preparation move forward in the background, then bring back a clear, specific action for approval. Really appreciate you calling this out—it’s a concrete example of what makes delegation feel trustworthy in everyday use.