Claude Code and Codex for Salesforce
The agent's pull request is on your desk. The Apex compiles and the tests pass, and you still can't tell if it queries inside a loop or ignores your sharing rules.
You work directly with an architect who can also write code. We're in Grünwald near Munich. More than 12 years on Salesforce, for companies directly and inside Salesforce partner teams.
How we work with AI coding agents
Context before code
The agents read the Jira ticket, the SharePoint document, and the Outlook thread that belong to the task.
A design your team decides on
The agents write an HTML document with a diagram. Finance, sales, and IT read it and decide, for example who approves an order over the credit limit.
A review by someone who knows the platform
We read every change for governor limits, sharing, and security. After more than 12 years on Salesforce we know where an agent goes wrong.
A pull request with tests
The agents implement what was decided. You get a pull request with passing Apex tests, ready for your own review.
Your rules, your tools
We use agents only as far as your policies allow. You approve the tools and the data access.
Our own app for both agents
We build HB Code, an app that runs Claude Code and Codex. It shows every command, file, and diff of an agent session.
Tell us about your project
Book a call, or send a few lines about your system and what should change.
Pick a time in our calendar.