Claude in Frankfurt: the Anthropic model and the people who build with it

Claude is a family of AI models from Anthropic, a company founded in San Francisco. Anthropic sells access three ways: the Claude apps for the browser, desktop and phone, an API that other software calls, and Claude Code for developers. The model line carries the names Opus, Sonnet and Haiku, with Opus built for the hardest reasoning work and Haiku for speed.

Two properties matter for finance work. Claude takes very long inputs in one go, so a whole credit agreement or a regulatory consultation paper fits into a single conversation without being chopped into pieces. And it can use tools: given access to a file, a database query or a web page, it fetches what it needs instead of guessing. That second property is what turns a chat assistant into an agent, which is the subject of AI agents in finance.

The work Claude takes over in a bank or an asset manager is document work first and code second. A compliance officer feeds it a BaFin circular and asks which of the firm's own policies the circular touches. A fund selector drops twenty factsheets in and asks for the differences in fee structure. An operations team turns a 40-page process description into a checklist the service desk can follow.

What all of these have in common is that the answer must be checkable. A finance team does not accept a summary it cannot trace back to a page, so the useful prompt asks Claude to quote the passage it relied on. That habit, more than any model property, separates teams that get value out of these tools from teams that quietly stop using them. The same discipline applies to every assistant on the market, which is why AI tools for finance compares them by the work they are given instead of by a score.

The regulatory side comes on top. A bank that puts a model into a process touching customers or credit decisions lands in the obligations of the EU AI Act, and the answer to "where does the data go" has to exist in writing before the first rollout.

Claude Code is a separate product for people who write software. You hand it a bug fix, a test to write or a migration that runs over several days, and it works in the repository: it maps and explains a codebase, follows imports across repositories, runs the tests and opens pull requests, as Anthropic's own product page describes it. It runs from a terminal, from VS Code and JetBrains, from Slack and from the web, on macOS, Linux and Windows, and it asks permission before it changes a file or runs a command.

For a Frankfurt bank's platform team the interesting detail is that it talks to the model API without a backend server of its own, and that Anthropic offers self-hosted environments which run inside a customer network. Claude Code in Frankfurt goes into what that means for a work machine behind a bank firewall.

Claude users in Frankfurt meet at a German-language meetup that the Agentic AI Community Frankfurt runs every other week at byte5 GmbH, Speicherstraße 1, about ten minutes on foot from the main station. The evening starts with a check-in, then guest talks, then a pizza break, and ends in smaller discussion groups. Subjects have included prompting, AI workflows, programming and web development, and recent editions covered memory in AI systems, holding context in Obsidian, and attacks on that memory. No technical background is required.

Finance Loop runs Claude Hacker House, the format for finance professionals who want to work with the tool instead of hearing about it. Claude meetups in Frankfurt describes both formats side by side and says which one suits which background.

Yes. The Claude Certified Architect Foundations exam tests whether someone can design a working system around the model instead of only writing prompts for it. The CCAF page covers the scope, who it suits and how the exam runs.

For teams the route is usually the other way around: a workshop first, a certification later, once people know which problems in their own department the tool actually solves. AI training in Frankfurt lists what the formats in the city cover.

Claude is one tool inside a city scene that also has universities, research institutes, bank innovation units and a row of meetups. AI in Frankfurt maps that scene by institution and event, the AI hub pages collect the meeting places, and AI in finance in Frankfurt covers the field inside banks and asset managers.

The technology underneath has its own page: large language models in finance explains what these models do and where they fail.