Specific Wants Coding Agents to Deploy Their Own Infrastructure

The Macro: Coding Agents Can Write Code but They Cannot Ship It

I have been using coding agents for about a year now. They are good at writing functions. They are decent at debugging. They are terrible at everything that happens after the code is written. Setting up a database. Configuring environment variables. Deploying to a server. Handling SSL certificates. These are all tasks that require interacting with infrastructure providers through dashboards, CLIs, and configuration files that were designed for humans, not for LLMs.

This is the bottleneck nobody talks about. The discourse around coding agents focuses almost entirely on code quality. Can the agent write a correct sorting algorithm? Can it refactor a React component? Can it fix a bug from a stack trace? These are solved problems at this point, or close to it. The unsolved problem is the last mile. You have working code on your local machine, and now you need it running in production. That step still requires a human.

AWS, Vercel, Railway, Render, Fly.io. All of these platforms were built for developers clicking through interfaces or typing CLI commands. Their APIs exist, but they were not designed with the assumption that the caller is an AI agent that needs to provision a Postgres database, configure a secret, and deploy a backend service in a single automated flow. The abstractions are wrong. The auth models are clunky. The error messages assume a human is reading them.

The bet that coding agents will eventually handle full-stack development, from writing code to shipping it, is a big bet. But it is the logical endpoint of where things are heading. If agents can write the code, they should be able to deploy it. The missing piece is infrastructure that speaks their language.

The Micro: A Stripe Tech Lead and an Energy Grid Engineer Built a Cloud for Agents

Specific is a cloud platform designed for coding agents to deploy code. Two commands define the entire workflow. specific dev creates a local development environment with all dependencies configured automatically. specific deploy pushes everything to production on their managed cloud platform. The agent handles the rest.

Fabian Lindfors is a founder. He was a tech lead at Stripe, where he worked on systems processing tens of billions in annual transactions. Iman Radjavi is the other founder. He spent four years building backend systems, including infrastructure managing national energy grids that process over a billion data points daily and healthcare systems under regulatory compliance. They are based in Stockholm and came through Y Combinator’s Fall 2025 batch.

The backgrounds matter here. Stripe infrastructure and national energy grid infrastructure are both environments where reliability is non-negotiable. You do not build systems at that scale without developing strong opinions about how infrastructure should work. The fact that both founders come from high-stakes backend engineering, rather than from the AI or developer tools world, is a signal about what they prioritize. This is an infrastructure company first, agent-friendly second.

The platform includes managed Postgres with vector and full-text search, S3-compatible object storage, Redis-compatible caching, cron job scheduling, secret management, frontend hosting with auto-generated domains and TLS, and a real-time sync engine. It is a full backend-as-a-service stack, not just a deployment target. The integration happens through an MCP server that connects to existing coding agents, so developers do not need to switch tools. They keep using whatever agent they already use, and Specific handles the infrastructure layer.

What I find compelling about this approach is the scope. Most agent-friendly deployment tools I have seen are narrow. They handle one thing well, maybe deployment, maybe database provisioning. Specific is trying to be the entire backend stack. That is ambitious for a two-person team, but the alternative is building something too narrow to be useful. If an agent can deploy a backend service but cannot set up the database it depends on, the automation chain is broken.

The competitive landscape includes Railway, which has good DX but was not built for agents. Render is similar. Fly.io has great infrastructure but the same human-first design assumptions. Neon and Supabase handle the database layer. None of them are positioning themselves as agent-native infrastructure. Specific is staking out that position early, which is either visionary or premature depending on how fast agent-driven development actually scales.

The Verdict

I think the thesis is right and the timing is early. Coding agents will need infrastructure that accommodates them, and the current cloud providers are not building for that use case. Specific is positioning itself at the intersection of two massive markets, cloud infrastructure and AI-assisted development, which is exactly where you want to be if the thesis holds.

At 30 days, I want to see how many agents have integrated through the MCP server and what the actual deployment success rate looks like. Getting a demo to work is easy. Getting a thousand different agent-generated codebases to deploy reliably is a completely different problem. At 60 days, I want to understand the economics. Cloud infrastructure is a margin business, and offering a full stack at a price that competes with AWS while maintaining reliability is hard. At 90 days, the question is whether developers trust their agents to deploy to Specific in production or whether it stays in the experimentation bucket. The gap between “I tried it and it was cool” and “my production app runs here” is where infrastructure companies live or die.