AGENT SWARMS ARE HERE
*To accelerate your business
Today we're launching Agent Swarms @grepdotai to accelerate your business by tackling the backlogs your team can't keep up with.
Swarms are teams of agents that work together to automate a task queue in parallel.
AgentRun is an open-source platform for supercharging your agents with self-improvement, durable execution and a single control plane.
Bring your agent and choose where it runs:
github.com/Parcha-ai/agentru…
Prediction: Sandboxes aren't a winner takes all market because agents will buy compute one tool call at a time.
A lot of the sandbox competition looks like a race to the bottom on price and startup time. The assumption seems to be that one provider wins the developer, then gets all their agent's work.
I don't think that's how this market will evolve.
One agent might need a cheap CPU sandbox to parse a file, a GPU for the next step, and a machine inside the customer's own infrastructure to touch sensitive data. Why would all three run in the same place?
Imagine a spot market for agent compute. Before a tool runs, the system finds the best available place to execute it.
First apply the hard constraints: which providers do I trust, where is this data allowed to go, what hardware does this step need? Then compare cost, availability and the time it will actually take to finish.
The cheapest hourly rate is useless if startup takes longer than the job. And sensitive data doesn't become okay to send somewhere because that provider is cheaper.
But what happens after the agent has done the same job a hundred times?
It learns the workflow. It should also learn where each step runs best: which environment needs a GPU, which can use a tiny container, which provider keeps failing on this workload, and which work must stay on your own machines.
That is where I think self-learning agents and sandbox routing meet.
We've already built some of the foundations in @agentrunai: durable execution, learning across runs, and a common computer layer with providers like Daytona and Modal.
My bet is there will be plenty of room for sandbox providers. They'll compete for particular workloads, and the agent will make the placement decision.
x.com/JoshARosen/status/2121…
Josh Rosen @JoshARosen Nov 15 Article The State of the Agent Sandbox Does every agent really need its own computer, or can much of its work be accomplished inside a far simpler execution environment? Over the past year, agent sandboxes have evolved from relatively
This is the exact pattern that @agentrunai uses. With durable agents the advantage is you can pick the right computer for each tool without wasting compute on the agent loop.
Josh Rosen @JoshARosen Nov 15 Article The State of the Agent Sandbox Does every agent really need its own computer, or can much of its work be accomplished inside a far simpler execution environment? Over the past year, agent sandboxes have evolved from relatively
AgentRun just passed 1,000 stars on Github!
Thanks to everyone who supported our launch this week!
Quick story on why we open-sourced our whole self-improving managed agent platform 🧵
And in 5 days, AgentRun passed 1,000 GitHub stars. Thank you to everyone who's tried it, shared it or contributed.
learn more: agentrun.ai
github: github.com/Parcha-ai/agentru…
AgentRun brings together everything we've learned from three years of building agents: self-improvement, jev workflows, durable execution, and a control plane into one easy-to-set-up repo.
We were also really inspired by YC founders like @james406 from PostHog who spoke at one of our YC dinners, @ishaan_jaff at LiteLLM which is a core part of our infra, and the team at @firecrawl.
All these products started as open-source projects.
We started with one piece: a small workflow language, or DSL, combining code and agent calls using @typesafeai Jev. We open-sourced the DSL in September, and the post went viral on X, but people were interested in the system around it too.
We built a new harness using @typesafeai's Jev that cuts the cost of repetitive work by 90%. The harness learns the job as it runs, moving steps from LLM calls to code.
Running 100,000 compliance alerts costs >$290K on Opus 5.
With agentrun() we got it down to <$26K.
Around the same time PG published his essay on making startups powerful.
That made us think: instead of asking for people's time and money to use our agent platform, what if we gave them something valuable for free?
paulgraham.com/powerful.html
The surprise was how much our control place was also a hit. Leaders at massive public companies wanted to know if they could use it with their own agents.
Prospects loved the self-learning system but wanted to run it in their own infrastructure. Our enterprise-specific harness wasn't the right fit for most of them. They already had agents and their own ways of working.
At YC's Early Access Program, a sort of demo day for customers, technical teams we met with were more interested in how we built our agents than in buying the product.
Our existing customers @grepdotai were growing >10% every week, but we kept coming back to the same question in office hours: how do we reach new ones?