The Agentic Engineering Company

From coding agents to software factories

We roll out, tune and run coding agents, harnesses and routers, so they cost less and ship better code.

Our team has built software and production AI for

trivagoSAPToniesEWEtecRacer360 Health & PerformanceConvointentFig Mayo

Where are your teams today?

We are your implementation partner at every stage: rollout, optimisation and the move to autonomous delivery, working inside your teams.

What agentic engineering means in practice

Agentic engineering is software delivery where engineers direct coding agents instead of writing every line themselves. Agents work with real repository context, and the changes they propose go through tests and review gates before they ship. Teams start with assisted coding, learn where agents are reliable, and then extend autonomy step by step. A software factory is the point where much of that loop runs on its own while people keep control of the decisions that matter.

We work out with your team which of these stages fits your codebase, people and constraints today, and build from there.

Traditional engineering

Engineers write and review every change by hand.

Rollout planAgent and model choice

Agentic coding

Agents work alongside every engineer, on a budget.

HarnessRoutingBudgets and capsTraining
Most organisations are here

Optimised

Every change to the setup is measured on your own code.

Benchmarks on your codeTask defaultsReview gates

Software factory

Agents take scoped work from issue to merge, with people in control.

Delivery loopRelease checks

Everything around the agent

Our engineers design, build and run it inside your infrastructure, with the models and tools you already use.

Lower cost

Model routingBudgets and hard capsLLM gateways and proxiesTask-type defaultsSpend attribution

Higher quality

Test and review gatesRepo context and rulesDocumentation rules

Rollout

Harness designCustom coding agentsHands-on training

Engineering work in public

These open source projects by Cloudsail's founder show the kind of engineering we bring, not results from client work.

Runmill on GitHub

Developer preview. Automatic merge modes are experimental.

Runmill takes an eligible Linear issue and hands it to Claude Code or Codex in an isolated workspace. Required checks run against the exact candidate commit, and a fresh agent context reviews the change. Runmill then opens the GitHub pull request and follows CI. Deterministic code, not the model, controls pushes, pull requests and merges.

ctxlane on GitHub

Tested against simulated vendor CLIs. Real accounts, keyrings and billing are not yet qualified for deployment.

ctxlane keeps personal, work and CI accounts for Claude Code and Codex apart. Each profile gets its own account state, and contexts can be bound to project directories. Static secrets managed by the wrapper stay in the native OS credential store. The official vendor CLIs still handle login and model requests. ctxlane is standalone and not integrated with Runmill.

What could you save?

Agent spend today

€600,000 a year

Illustrative savings at an assumed 19 percent

€114,000 a year

Check against your real usage

An illustrative scenario, not a measured average or guaranteed outcome. A usage review tests the assumption against your actual costs and quality.

Measured on your code, not on benchmarks

We replay your merged changes on the current setup and the alternatives. A saving only counts when quality holds.

Spend per finished task

Illustrative example, synthetic data

Bug fixes
€3.10 to €2.30
Refactors
€4.20 to €3.30
Tests
€1.90 to €1.30
Reviews
€1.10 to €0.85

€62000 to €49000 a month

Rework 11% to 10%

Built for the people who own the result

CTO

Agent spend you can explain, and evidence for every change.

CIO

One governed setup across teams, vendors and approved models.

Head of AI or Core AI

Tested defaults for every model in your catalog.

Head of engineering or platform

Agents your engineers trust, use well and keep within budget.

Your code stays yours

We are independent of model and tool vendors. No vendor pays us, so our recommendations follow what works for you.

We work inside your environment.

A data processing agreement comes with every contract.

Reporting by team, never by individual developer.

Works council requirements handled early.

Questions

What does the usage review cost?

Nothing. You share one month of usage data, and we return a one-page breakdown of where the spend goes.

We cannot share code with an outside party.

You do not need to. We work inside your environment under a data processing agreement, and nothing leaves without your written agreement.

We already work with our vendor's engineers.

They help with their own tool. We compare tools and models on your code, and route work to whatever performs best for the money.

How much of our team's time does this take?

An hour for the usage review. For a three-week engagement, one engineering contact per team and 30 minutes a week with the owner.

Do you guarantee savings?

No. We measure savings on your own work and say how certain each figure is.

Request a usage review

Share one month of usage data. We return a one-page breakdown of where your agent spend goes and where it can come down. No charge.

This opens your email app. Review and send the draft there to request your usage review. You can also email miki@cloudsail.com directly.