Asher Dale

founding engineer @ (YC S25)

[Asher Dale @AsherDale13]

@ishaansehgal @omnaraai 🌎

[Asher Dale @AsherDale13]

a quick look at how @omnaraai handles sandboxes and BYO machines

[Asher Dale @AsherDale13]

@omnaraai 🚀

[Asher Dale @AsherDale13]

@kartiksmath @omnaraai greatest name in the world 🙌

[Ishaan Sehgal @ishaansehgal]

first office! 💪

[Ishaan Sehgal @ishaansehgal]

you can now see exactly what every org, project, profile, and agent is spending in @omnaraai, subagents included

[Asher Dale @AsherDale13]

@csparks1919 @omnaraai huge that you can define custom subagents instead of being restricted to only inheriting the parent 🚀🚀

[Christian Sparks @csparks1919]

Gonna be posting some demos here now! Here's a quick little demo of subagents in the new https://t.co/Yp1nDGuP75

[Asher Dale @AsherDale13]

@kartiksmath @omnaraai just another day at the office 🫡

[Asher Dale @AsherDale13]

building out inter-agent memory for cloud agents on @omnaraai

[Ishaan Sehgal @ishaansehgal]

Omnara is now available over MCP

MCP joins the @omnaraai API, SDK, and CLI. Connect Claude Code, Codex, Cursor, or any MCP client and spin up production-grade agents

[Ishaan Sehgal @ishaansehgal]

OpenRouter cloud agents this week:
#11 AI for Roblox games
#14 a betting app
#15 Omnara

3.68B tokens. we'll take it

[Asher Dale @AsherDale13]

@ishaansehgal Next level stuff 🔥🔥🔥

[Ishaan Sehgal @ishaansehgal]

Introducing Omnara: the open-source alternative to Claude Managed Agents.

Omnara is an API for creating and interacting with production-grade agents. You build the agent. Omnara handles hosting, durable state, sandboxes, permissions, secrets, and the rest of the production infrastructure.

Available now @omnaraai

[Asher Dale @AsherDale13]

@MatthewBerman @bot this would be awesome 🔥

[Asher Dale @AsherDale13]

@kartiksmath insightful thinking. useful to read for people that are over-indexed on harness optimization with diminishing returns

[Kartik @kartiksmath]

Just use Pi, bro https://t.co/yvScQj4p0r

[Ishaan Sehgal @ishaansehgal]

your agents can now wake themselves up at 6am and get to work without you

whatever config, models, tools, MCPs, skills, or setup you want, running on a schedule

we've been dogfooding it for our daily usage reports @omnaraai

managed agents on your schedule

[Ishaan Sehgal @ishaansehgal]

good question. omnara sits a layer below eve

eve is an application framework, you author an agent as a typescript codebase and ship it as a deployment. in omnara agents are treated as durable api objects you can launch and steer in real-time

this makes spinning up agents easier and more powerful. for example spin up agents per customer, per ticket or per slack thread with an api call. change an agent’s behavior by posting a new config (not a new deploy). let non-engineers launch and steer agents from the console. rbac, per-project grants on models, secrets, and machines, and spend limits out of the box instead of building them. and every agent's history lands in your own postgres, queryable for any purpose

omnara is also vendor-neutral - bring your own model keys, pick your sandbox vendor (e.g. blaxel, daytona, unikraft) with your own account, or attach your own hardware (your laptop counts). eve's happy path is vercel-shaped: ai gateway, vercel workflow, vercel sandbox

to be fair, eve is a great fit if you're building one deeply integrated agent inside a typescript app (custom channels, react hooks)

whereas omnara lets you run agents as a service you call. this gets into a lot of the primitives we built to make agents flexible and powerful like the control plane, the log is the agent, stateless workers advancing the loop, machines are tools, and orgs, projects, roles, and grants govern everything

[Ishaan Sehgal @ishaansehgal]

What is a "managed agent"?