About

Production AI engineer. Shipping MCPs + Aufgaard plugin for production AI ops.

I ship production-AI ops tools.

The shipped artifacts:

7 MCP servers — silentwatch, bash-vet, output-vetter, skill-vetter, cost-tracker, health, upgrade-orchestrator. All MIT, all on PyPI.
1 plugin — Aufgaard — bundles all 7 + skills + hooks for Claude Code and Cursor. Submitted to Anthropic and Cursor marketplaces.
1 bundle — Production-AI MCP Suite — buyer-only field reference, $99.
1 Notion template — AI Production Discipline Framework — the 14-pattern catalog these MCPs implement, $29.
1 free Custom GPT — AI Production Auditor — paste your config, get a 5 Cs audit report.
Long-form writing — Writing →

The methodology behind the tools comes from cataloging public production-AI incident reports — 200+ threads on HN, Reddit, X over the past year. The patterns are remarkably consistent across stacks (LangChain, LlamaIndex, vanilla SDK, custom agent harnesses), audiences (B2B SaaS, consumer features, internal tools), and scales. The 35-pattern catalog organizes them.

The 7 MCP servers each address one or more patterns from the catalog inline — sub-second, local, free, MCP-native — designed to be called from Claude Code / Cursor / Cline / OpenClaw agents during the conversation, not as separate eval-pipelines.

What I’m available for

Production AI audits — $1.5–7.5k tiered, 1-3 weeks. Outside-practitioner scoring against the 35-pattern catalog.
Custom MCP server builds — $8–45k tiered, 1-3 weeks. Adapt the templates to your stack — different harness, custom logging schema, vendor billing API, internal compliance.
Advisory — short-form consulting on production-AI architecture decisions.

Hire me →

What I’m not

Not an AI influencer. The tools speak.
Not a SaaS founder. Audits + custom builds, not subscriptions.
Not a course creator. The 35-pattern catalog is the methodology; the MCPs implement it.

Contact

Email: hello@temhan.dev — reply within 2 business days

Other surfaces:

Based in London / Karachi.