Aident AI Beta 2 - Open-world automations, managed in plain English
by•
Meet the future of work—AI that actually runs with you. Build and manage open-world automations in plain English across Discord, Slack, X, Shopify, and more with 1000+ integrations, 23000+ actions, and 1000+ templates. Trigger on real-world events, get updates in your favorite chat apps via IM + MCP, and monitor runs, approvals, and issues from one live dashboard.
Replies
Trufflow
I'm not always the best at prompting so I always appreciate a "Plan" mode to get me started. Is there something similar in Aident so that the automation that gets set up is what I intend it to be?
Aident AI
@lienchueh Yes! Actually, our agent is default to a "plan" mode where it will gather enough information before getting started into building an automation playbook. After creating your playbook, you can always "test-run" it then come back to the playbook editor to prompt agent to fix.
Really interesting framing! I’m curious who you see as the ideal user for Aident right now. Is this primarily built for non-technical operators who want automation without thinking about logic, or more for technical teams who just want to move faster? Also wondering where you see the main difference vs tools like Zapier, Make, or newer AI automation layers. Is the key advantage the instruction-first interface?
Love the design perspective you’re bringing into automation!
Aident AI
@tereza_hurtova We are seeing more of the prior - non-tech operators are the real source of truth when it comes to the day-to-day business workflows, its their life so they know better than anyone else. we want to empower them to feel the same magic we have been experiencing with Claude Code, Codex and OpenClaw with their real work, real results and real wows.
Regarding the differentiator vs existing tools, we call ourselves the next-gen automation platform, cuz we have changed the games of the old way of automation to determine the every single step of automation in an SOP manner into a more goal driven automation in the form of natural language documents. It is game changing cuz we are abstracting up so that people can think more about what to do and automate rather than how exactly the task should be done and stuck in details.
Congratulations
Aident AI
@madalina_barbu thanks!
Copperlane
Looks cool! Curious how Aident handles reliability when automations get complex across many tools. If a step breaks or an API changes, does the Playbook automatically adapt or flag the issue?
Aident AI
@brianna_lin Great questions! We are internally testing 2 things: 1. Aident has a tool to directly open issues and assign to us - we make it to debug & implement fairly soon. 2. Aident has a tool to improve the playbook, with approval from users.
Zawa
curious about the "open-world" aspect here. most of these natural language automation tools tend to break once you step outside a very rigid happy path. since you're supporting 23k+ actions, how do you handle state management when one of those mid-stream actions fails or returns an unexpected schema? i'd love to know if there's a way to inspect the underlying logic tree or if it's all just black-box magic.
Aident AI
@tony_hsieh2 Great question! Under the hood we are cooking both passive and proactive ways to gather feedbacks for agents to use these actions - passive metrics are statics signals like success rate, response time, latency, etc; proactive feedbacks are agent rating and commenting after the action usages (like agent Yelp for the services/actions). We haven't shared much about these little magics yet as we are still polishing these to prepare a much larger scale in our next version. So, it's a little secret of our product and a little "secret" between us haha! Hope it answers your questions (i know i intentionally skipped the part about how to handle errors but as we know which ones are more reliable we know what to expect when it fails). :D
Told
The plain English approach to automation is a really smart unlock — most no-code tools still require users to think in logic trees. Curious how you handle ambiguity when a user's instruction could map to multiple action chains? That edge case is where these tools usually break trust. Would love to see how the error handling surfaces back to the user when something unexpected happens in a real-world trigger scenario.
GPT Nitro for Github PR
How does this work with OpenClaw/ do it compete?
Aident AI
@miley_fu you can install aident as a skill in OpenClaw
Wow! this looks like an all in one tool that I can use for my marketing!
Aident AI
@abhinavramesh glad it'd be helpful! let me know how you like it :D
Me describing my automation to Aident: "Do the thing like Zapier/n8n but don't break every Tuesday" 🙏
Aident AI
@ilya_lee lol
Aident AI
@ilya_lee my prompt to claude code "claude --dangerously-skip-permissions, but don't do dangerous things"
Its amazing how the world of automation software has improved over the last few years - between manually stitching them together in platforms like Zapier to using AI to figuring it all out.
Very cool guys!
Aident AI
thanks @jake_friedberg ! it's next-gen automation haha!