Instruct - Automate anything with a single prompt
by•
Tired of complex automation tools? Instruct is the first truly no-code platform where you can build, edit and run powerful AI agents using only natural language - just explain your task, and let Instruct handle the rest.
Replies
GraphBit
Really like the way you framed agents vs workflows. That distinction hits home, workflows break at the first edge case, while agents adapt. Making that power accessible to non-technical users feels like a big unlock. Curious: how do you balance flexibility with keeping users in control (so they trust the agent won’t go off-track)?
Instruct
@musa_molla Thank you, Musa! That’s exactly the gap we’re trying to close 🙂 We give agents a lot of autonomy under the hood, but we surface their reasoning and next steps in plain language so users can step in, edit, or override at any point. In practice it feels like “guided autonomy” — you still get the flexibility of an agent but always know (and can change) what it’s about to do.
Instruct
@musa_molla Thanks Musa, we think that this is really important, too. There's a lot to do in the space around balancing flexibility with user control and we have a lot planned! Fundamentally, we also believe that as the models continue to improve, the likelihood that they go off track will decrease.
Congrats on the launch, Matt! Love the shift from rigid workflows to adaptive agents , that’s a real game-changer. Curious though, how are you planning to handle early user feedback at scale? That’s usually a pain point for fast-growing teams.
Instruct
@gloria_anikpe Thanks Gloria! We're trying to talk to as many customers as we can right now - and implement their feedback.
Selene by Atla
Very excited for this! I haven't found a good automation tool for AI workflows so far. So much potential here
Instruct
That's great to hear@maurice_burger! Please do let us know how you find using Instruct :)
Verte
niceeee congrats!
Instruct
@ivan_porollo Thanks Ivan! Hope you're well!
The product seems really great and handy, but I just wanted to know how it differs from COMET by Preplexity.
Instruct
@headsin Thanks, we're users of Comet internally - and that's great for those use-cases that happen infrequently. For common, repetitive work, we think Instruct is really powerful!
Recall.ai
Congrats team!!
Instruct
@amanda_zhu Thank you Amanda, we really appreciate this :)
Instruct
Thanks@amanda_zhu- appreciate the support!
Congrats on the launch! Can we chain multiple agents or tasks into complex workflows, or does each prompt yield only a single, isolated outcome?
Good luck on the launch @mattfalconer I'd like to ask that, are you dependent on any LLM service or do you have an agnostic structure?
Instruct
@savaskalkan Behind the scenes, we use a collection of models to provide the experience you get with Instruct - we may look to open this up to the user in the future. Is this something you might be interested in?
Instruct
@mattfalconer @savaskalkan Hey Savas, we're not tied to any particular LLM service. We're continuously evaluating the latest models available to ensure our users get the best performance possible.
Raycast
I couldn't find any information about your pricing. How much does this cost?
Instruct
@chrismessina Hey Chris! Pricing is pretty new for us. We current have a credits system, it's one credit per minute that your agent runs. We're working to improve this!
I recently tried Instruct and I have to say it's very impressive! The agent creation is absolutely spot-on and I managed to build my first LinkedIn post agent, which worked great. However, one thing to note is that it consumed 50 credits for a single run, which feels a bit expensive for frequent usage.
I've used multiple automation platforms, and I think the credit policy could be improved. Currently, you provide 50 credits per month, but since a single automation run can consume all 50 credits, I'd recommend offering 50 credits per day for at least the initial few months as a trial or testing period. This would allow users to test and experiment more freely with building agents. Later, you could upgrade the pricing plan as needed based on usage patterns. Just a friendly suggestion—ultimately it's your decision. Rest, it's been a great experience.
Instruct
@abhishek_mishra74 Thanks Abhishek, I agree that we need to change the credit model to something that's more frequent to ensure users can test properly. We're working on it, stay tuned :)