Deutsch

Claude Managed Agents and Waterglass Waves compared

Both platforms follow the same idea: you define AI agents, the platform runs them. The differences are in the AI models, in where processing happens – the USA or the EU – and in who you contract with.

Create a Waves account Learn more about Waves
Introduction

What are managed AI agents?

Managed AI agents are agents you only define, not operate: the platform takes care of isolated execution, schedules, credentials and logging. Anthropic popularised the concept with Claude Managed Agents.

Technically the two platforms overlap a good deal – isolated execution environments, encrypted vaults for credentials, versioned agent definitions as YAML. The differences sit elsewhere: which AI models you can pick, where processing happens, who you contract with, and how broad the feature set is.

Comparison

Waves and Claude Managed Agents in detail

Criterion Waterglass Waves Claude Managed Agents
AI models The model is chosen per agent and can be overridden per session Claude models
AI model hosting In Europe or the USA, depending on the model you choose USA
Where processing happens Nuremberg (Germany) Anthropic's own infrastructure in the USA (no EU region documented)
Contracting entity Waterglass (Austria) Anthropic (USA)
Execution environment One isolated environment per session. Network access controlled via allow-lists One isolated environment per session. Network access either open or restricted via allow-lists
Credentials Vaults, encrypted with AES-256-GCM. Values are write-only and never appear in any response or log Vaults. Values are write-only and are substituted only on the outbound call, never visible inside the execution environment
Agent definitions Versioned definitions, editable as YAML in the console or through the REST API Versioned objects with immutable versions, editable as YAML via the command line (CLI) or REST API
Running on your own infrastructure No: Waves runs entirely on European infrastructure, operated by Waterglass Partly: tool execution can run in your own infrastructure, while the agent loop and model calls stay with Anthropic
Feature scope Isolated execution, schedules, skills, persistent memory, MCP connections, browser access and a record of every action The same core, plus multiple agents working together and grading results against a given rubric
Onboarding Sign up yourself or get a personal introduction Sign up yourself
Billing €29 per month net on the Basic plan, plus model token consumption paid out of credit Usage-based on model token consumption
Sources and date

As of August 2026. Statements about Claude Managed Agents – also widely called "Anthropic Managed Agents" – follow the public documentation at platform.claude.com/docs; statements about Waves reflect its current feature set. We correct this table when anything changes – please send corrections to hi@waterglass.ai.

Summary

When to pick Waves, when Anthropic?

Claude Managed Agents fit when …

  • you are on Claude anyway
  • you need multiple agents working together or automatic grading against a rubric
  • processing on Anthropic's infrastructure is acceptable for your use case

Waterglass Waves fits when …

  • EU data residency is a requirement
  • you want to pick and switch AI models freely
  • a European contracting entity is required
  • you would like a personal introduction
Questions & answers

Common questions about Waves and Anthropic

Is Waves an alternative to Claude Managed Agents?
For many use cases yes. Both platforms cover the core: define agents, have them run in isolation, manage credentials, log actions. Waves offers free model choice, EU data residency and a European contracting entity. Which side wins depends on your requirements. We are happy to work through that in a call.
Can I use Claude models in Waves too?
Yes. Claude is one of several providers in Waves, alongside GPT, GLM, Kimi and Mistral. You pick the model per agent and can switch it anytime. New models are added continuously.
Can Claude Managed Agents be run in Europe?
Partly. Anthropic offers self-hosted execution environments: tool execution – shell commands, file access, code – then runs in your own infrastructure. The agent loop and the model calls still stay with Anthropic. If your requirement is that all processing happens in the EU, that is not sufficient.
What does EU data residency mean for Waves in practice?
The platform runs in a data center in Nuremberg, including object storage – so compute and data residency are within the EU. For the AI models, Waves shows each provider's jurisdiction and you decide which model an agent may use. We work through your specific requirements during onboarding.
What does the difference mean for the GDPR?
The GDPR does not prohibit processing outside the EU, but it attaches additional requirements to it, among them a legal basis for the third-country transfer and a data processing agreement. With Waves, compute and data residency sit in a data center in Nuremberg and the contracting entity is an Austrian company. A third-country transfer only arises where you allow an agent to use a model outside the EU. For Claude Managed Agents, no EU region is documented. Which requirements apply to your use case is something we work through during onboarding.
How can I try Waves?
You can start self-serve or have us guide you. Enter your email address on the Waves product page: we get back to you within 1–2 business days and, if you like, set up your organization and define the first agents together with your team.

Try Waves yourself

Create an account, claim €10 of starter credit and define your first AI agent. Basic is €29 per month net.

Create account

Not sure which fits?

Book a short call. We will tell you honestly whether Waves fits your use case – including when the answer is no.