Hoplite (YC S26) builds cloud agents for programming in one minute
Hoplite, a Y Combinator S26 startup founded by Ryan Morrissey and Bence Redmond, launched a platform Tuesday that lets developers set up cloud coding agents in one minute by transferring their local sessions and MCP servers. Source installed it last week and saw immediate results with proactive PRs from Sentry logs.
Context from: Hoplite | Ycombinator
The decision it puts on your desk
engineering teams that are already seeing value from local coding agents but have not moved them to the cloud are leaving throughput on the table. Set up Hoplite at hoplite.sh, transfer one active agent session, and measure whether cloud execution meaningfully changes your merge cadence within two weeks.
In the next ten years, the best software startups will be those that can effectively deploy tokens.
That is the bet Hoplite is making. Ryan Morrissey and Bence Redmond met in college in Ireland and spent two years building projects and doing freelance work together until they got sick of building other people's visions. Bence interned at AWS. Ryan interned at Stripe. Having seen how much faster cloud agents let engineering teams move, they dropped out to build this for everyone.
The problem Hoplite solves is one that anyone who has run multiple coding agents locally understands immediately. Cloud coding agents are painful to set up. They lack ergonomics. They do not let you move as fast as you could.
Hoplite transfers all of your local sessions and MCP servers to the cloud, so you pick up right where you left off. It also configures every dependency and CLI your projects need, letting you see instant previews of what your agent is doing.
The result is a software factory for businesses that want to deploy more tokens but do not have the time or resources to build the infrastructure themselves.
Source installed Hoplite last week and saw results immediately. Their Sentry logs went from largely ignored to generating proactive PRs the moment something breaks. Iteration cycles are significantly faster, and lower-priority tickets are getting merged continuously.
The company frames this as the infrastructure layer that separates the startups that can effectively deploy tokens from those that cannot. I find that framing useful. The conversation about AI in software engineering has been dominated by model quality, but model quality only matters if you can actually run the agents somewhere. For most teams, that is still a manual, fragile, and slow process.
Hoplite is essentially saying the bottleneck has shifted. The models are good enough. The infrastructure to run them is not.
The founders are betting that as more companies try to go from one engineer-plus-Copilot to multiple autonomous agents running in the cloud, the infrastructure layer becomes valuable in its own right. Not because it is hard to build once. Because it is hard to keep running, keep secure, and keep fast across projects, dependencies, and teams.