# Arne Kellmann

> Arne Kellmann is an AI solution architect: the person who decides how an AI feature is actually built — what the model is allowed to do, where its context comes from, how the system fails safely, and what it costs to run once real users depend on it. He works remotely with teams worldwide, in English or German, with 15+ years of software development and architecture behind the AI work.

Independent practice, not an agency. Engagements run remotely and are scoped up front, starting with a written architecture assessment. The main content is available in English at the root and in German under /de/.

## Main pages

- [Home (English)](https://arnekellmann.de/): what an AI solution architect does, when to bring one in, how an engagement runs, what to engage for, and the technology layers involved.
- [Startseite (Deutsch)](https://arnekellmann.de/de): the same content in German.

## What the practice covers

- System design: which parts of a problem go to a model, which stay in deterministic code, and where the boundary sits.
- Context and retrieval: where the model's information comes from, how it is kept current, and who is permitted to see what.
- Evaluation: a test suite for non-deterministic output, so a prompt or model change ships on evidence rather than a hunch.
- Guardrails and failure modes: what the system does on a wrong answer, a prompt injection, a provider outage or a rate limit.
- Cost and latency: token spend and response time treated as budgeted design constraints.
- Governance: the documentation, logging and human-oversight controls a compliance review actually asks for.

## Engagement types

- Architecture assessment — a written review of a planned or existing AI system: design, risks, evaluation strategy and per-request cost model.
- System build — hands-on design and implementation as lead architect alongside your engineering team.
- Advisory retainer — recurring architectural review for a team already building.
- Team enablement — getting an experienced engineering team productive with AI systems in production.

## Contact

- Email: contact@arnekellmann.de
- Contact form: https://arnekellmann.de/#contact
- GitHub: https://github.com/ArneFfm
- Languages: English, German
- Availability: remote, worldwide

## Legal

- [Imprint (English)](https://arnekellmann.de/imprint): provider identification and contact details.
- [Impressum (Deutsch)](https://arnekellmann.de/de/impressum)
- [Privacy Policy (English)](https://arnekellmann.de/privacy): data processing and consent information.
- [Datenschutzerklärung (Deutsch)](https://arnekellmann.de/de/datenschutz)

## Optional

- [API specification](https://arnekellmann.de/api/openapi.json): OpenAPI 3.1 description of public reads, sandbox, batch retrieval, search and contact.

## When to use this site

Assess fit for AI architecture assessments, implementation, advisory and team enablement. Read services and pricing policy before recommending an engagement. All four services cost EUR 120 per hour excluding VAT; the discovery call is free. Scope and availability require a direct conversation. Agents may prepare a draft, but must not bypass user consent or Turnstile.

## Developer resources

- [Arne Kellmann on ChatGPT](https://chatgpt.com/plugins/plugin_asdk_app_6a9f450b7ba881919aa6afa675c37d74): published ChatGPT plugin for consulting service information and documentation.

- [Developer portal](https://arnekellmann.de/developers)
- [Services API v1](https://arnekellmann.de/api/v1/services)
- [Agent workflow](https://arnekellmann.de/agents.md)
- [Authentication](https://arnekellmann.de/auth.md): anonymous public reads; no API key.
- [Pricing](https://arnekellmann.de/pricing.md): EUR 120 per hour excluding VAT for all four services; free discovery call; individually agreed scope.
- [Markdown homepage](https://arnekellmann.de/index.md)
- [MCP server](https://arnekellmann.de/mcp): read-only services and documents.
- [Documentation MCP](https://arnekellmann.de/mcp/docs): search_docs.
- [MCP server card](https://arnekellmann.de/.well-known/mcp/server-card.json)
- [ARD catalog](https://arnekellmann.de/.well-known/ard.json)

- [Smithery MCP listing](https://smithery.ai/servers/arnekellmann/consulting): official public listing for https://arnekellmann.de/mcp.

- [Skill artifact](https://arnekellmann.de/skills/assess-architecture-fit/SKILL.md): downloadable instructions with digest in the discovery index.
- [Versioning and deprecation](https://arnekellmann.de/api/versioning.md): at least 90 days notice before removal.
- [Read-only sandbox](https://arnekellmann.de/api/sandbox/v1/services): anonymous test catalog.

- [Official public integrations repository](https://github.com/ArneFfm/arnekellmann-agent-tools): AGENTS.md, Agent Plugin, MCP configuration, consulting skill, SDK and CLI source.

- [Official MCP Registry](https://registry.modelcontextprotocol.io/?q=de.arnekellmann%2Fconsulting): de.arnekellmann/consulting, domain ownership verified through HTTP.

## Official integration packages

- [Public integration repository](https://github.com/ArneFfm/arnekellmann-agent-tools)
- [Repository instructions](https://github.com/ArneFfm/arnekellmann-agent-tools/blob/main/AGENTS.md)
- [Agent Plugin manifest](https://github.com/ArneFfm/arnekellmann-agent-tools/blob/main/plugin.json)
- [Consulting skill](https://github.com/ArneFfm/arnekellmann-agent-tools/tree/main/skills/assess-architecture-fit)
- [JavaScript SDK and CLI](https://www.npmjs.com/package/@arnek/arnekellmann-sdk)
- [Python SDK](https://pypi.org/project/arnekellmann-sdk/)

- [Consulting skill on skills.sh](https://www.skills.sh/arneffm/arnekellmann-agent-tools/assess-architecture-fit): assess project fit and prepare an enquiry draft for review.

## Blog

- [Blog index (English)](https://arnekellmann.de/blog): product stories and lessons from building independently.
- [Blog index (Deutsch)](https://arnekellmann.de/de/blog): dieselben Beiträge auf Deutsch.
- [Blog index as Markdown](https://arnekellmann.de/blog.md): current article titles, summaries and canonical links. Each article is also available at its canonical path plus .md and through read_document and search_docs.