The Agentic Engineering Company
Roll out coding agents without giving up control of cost and code
We roll out coding agents in your engineering teams and train everyone who works with them. We also build the platform underneath, in your cloud or your own data centre, with a budget for each team.
Every step is measured on your own repositories, and review and release stay with your engineers.
Works with
Claude CodeOpenAI CodexCursorGitHub Copilot
No model or tool vendor pays us, so our recommendations follow your results.
Low risk from the first call
- The first call is free and takes 30 minutes.
- After it, you get a fixed price for each step in writing.
- You can stop after any step and keep what has been built so far.
Our team has built software and production AI for
What you get
What changes for your teams
You can start with any one of these.
Also for engineering teams
Harness engineeringEvaluations on your repositories
All coding agent servicesEvery team works the same way
One shared setup instead of a different tool for each engineer.
Your people know how to work with agents
Training on your own code and tickets.
Your code stays where you decide
A platform in your cloud or your own data centre.
A bill you can explain
Spend by team and task, with a budget for each team.
The platform
The platform under your coding agents is yours
We build it in your environment and hand it over. Nothing ties you to us or to one vendor.
Observability and cost
Spend per team, cost per merged change
People
Training for engineers, product managers and domain experts
Verification
CI gates, evals on your own repositories, AI first review, human merge
Team knowledge
Rules, skills, MCP catalogue, architecture decision records
In your cloud or data centre
Agent harness
Claude Code, Codex, Cursor, Copilot, hooks
Sandboxes
Isolated containers, outbound allowlist, short-lived credentials
AI gateway
LiteLLM, keys and budgets per team, MCP gateway
Models
Hosted in the EU, or open-weight on your own servers
Governance and compliance
Works council, DPA, AI Act
Would you like to talk this through for your own case? The first call is free and takes 30 minutes.
Book a 30-minute callSoftware Manufactory
Domain experts and engineering build software together
Agents do the work and your engineers sign off every change. Each stage is measured on your own code before the next one begins.
Autonomous software factory: what it takes and where its limits are01Assisted coding
An engineer and an agent on one task
02Supervised agent tasks
Agents take whole tickets
03Background agents
Narrow work runs unattended
04Software factory
From issue to merge on a written delivery line
05Domain experts join
Product managers, designers and domain experts work on the same path as your engineers
Self-check
Where are you today?
Choose the line that fits your teams and see what the next stage needs.
01Your next stage
Assisted coding
One team works with a shared setup and is trained on its own code.
Agent harnessPeople
02Your next stage
Supervised agent tasks
Your rules live in the repository, and every change passes the same checks before an engineer merges it.
Team knowledgeVerification
03Your next stage
Background agents
Agents run in sandboxes, and every team has its own keys and budget.
SandboxesAI gateway
04Your next stage
Software factory
Evals on your own repositories decide what passes, and you see the cost of every merged change.
VerificationObservability and cost
05Your next stage
Domain experts join
Product managers and domain experts are trained, and it is clear what they may change on their own.
PeopleGovernance and compliance
After the first call you get a written proposal. You can stop after any step and keep what has been built so far.
- 01
Your environment, your rules
Every data flow is agreed with you before work starts.
- 02
A data processing agreement in every contract
With the documentation your data protection team needs.
- 03
Teams and tasks, never individuals
Works council requirements are handled before anything goes live.
- 04
Everything we build stays yours
Code, test sets and runbooks, written for your team to run.
- 05
Training that counts towards the AI Act
Article 4 asks for AI literacy among staff. Training on your own code is one such measure.
Clients
What clients say
“They were embedded in our team, which was really effective.”
“We are impressed with their engineers' high-quality work.”
“Their responsiveness, friendliness, and successful project delivery were top-notch!”
Beyond engineering teams
All servicesThe same platform also carries AI for your business processes.
Ask AI for a summary of Cloudsail
Pick a topic and open it in your assistant.
Summarise what Cloudsail (cloudsail.com) does, which services it offers and how a first project with them runs.
FAQ
Common questions
What do coding agents cost per developer, and can we cap it?
It depends on usage more than on seats. Anthropic reports $150 to $250 per developer per month for Claude Code in enterprise deployments, and all four tools let admins set spend limits.
How do we prove that coding agents pay off?
Agree the goals before the trial starts, then compare the agents' changes with your team's own on real tickets. The free AI usage review gives you the starting figures.
Does our code leave the company, and is it used to train models?
With hosted tools, prompts and file contents go to a model provider. On business plans the four vendors state that they do not train on your code by default. Retention and region depend on your plan.
Can people without an engineering background really build software with coding agents?
For prototypes, internal tools and small changes, yes, inside the same setup as your engineers. An engineer reviews every change before it merges.
Who reviews all the code that agents write?
An AI reviewer can take the first pass. The merge decision stays with a named person other than the engineer who started the agent.
Is code written by agents secure?
Only as far as your checks make it so. Give the agent narrow, short-lived credentials, and keep branch protection and human review out of its reach.
Do we need to involve the works council before rolling out coding agents?
In Germany, often yes, wherever agent logs or usage data could show how individual employees behave or perform. Involve the works council early and agree what is logged and who sees it. This is practical guidance, not legal advice.
Which coding agent should we choose, and how do we avoid lock-in?
Run the candidates on the same tasks from your own repositories and compare accepted changes, review time and cost. Keep the rules for agents in your own repository, so they move with you.
8 more questions
Will coding agents work on our legacy codebase?
Often, once a fresh checkout installs, builds and tests with commands the agent can read. Where tests are thin, start with narrow tasks.
How do we roll out coding agents without a mandate that backfires?
Start with one team that wants to try them, and agree what will be measured before the pilot begins. Measure by team and task, never by individual.
What may agents touch, and who decides?
Your team decides, and the rules are enforced outside the agent. Typical limits are protected paths, no production secrets and a required human approval before a merge.
Does the EU AI Act require us to train our developers?
Article 4 obliges companies that use AI to ensure enough AI literacy among their staff, without prescribing a format. Training on your own code is one such measure. This is not legal advice.
Can we use coding agents without our code leaving our network?
Yes, with open-weight models on your own servers behind a gateway. They are usually weaker on hard tasks than the best hosted models, so a common setup combines both.
What is an autonomous software factory, and do we need one?
A delivery line in which coding agents take scoped work from issue to merge, while people decide what gets built and what may merge. It comes after your teams already work with agents on a shared setup.
How is this different from a large integrator or a tool vendor?
An integrator builds your software with its own agent teams on its own platform. A vendor sells you one tool. We make your own teams able to work with agents, on a platform that is yours.
What does Cloudsail do?
We roll out coding agents in engineering organisations and build the platform underneath, so cost and code stay under the customer's control.
Start with one team
Bring the state of your rollout, or a bill you cannot explain. You talk to an engineer for 30 minutes, and we do not need access to your systems.
Mickey Graf
Founder and CEO of Cloudsail
Prefer to write first? Leave your email and a few lines, and we reply with times.
Look for the draft in your email program
Send it from there, and Mickey replies in person with times for a 30-minute call. If no draft has opened, write to us at miki@cloudsail.com.