DT Technical Specialist

A Claude Desktop project providing a Dynatrace Technical Specialist AI agent. Configure it once in Claude Desktop and the agent answers Dynatrace, Bindplane, Bluebox, and OpenPipeline questions with the precision and guardrails of a senior field engineer.

How it works

This repo contains no executable code — only curated Markdown documents used as context in a Claude Desktop project. Instructions.md defines the agent's persona, question routing logic, and guardrails. The seven reference files are its knowledge base.

Claude Desktop projects let you attach both a system prompt (Project Instructions) and a set of knowledge files to a persistent conversation context. Instructions.md goes in as the Project Instructions; the seven reference files are uploaded as project knowledge. Every conversation in the project starts with all of this context already loaded.

Audience

Dynatrace Technical Specialists, field engineers, pre-sales engineers, and support practitioners who need fast, accurate answers during customer engagements, proof-of-concepts, or troubleshooting sessions.

Setup

You need Claude Desktop with a Pro or Team plan (project support is not available on the free tier).

1. Clone this repository:

   git clone https://github.com/virtualrussel/DT_Technical_Specialist.git
   

2. In Claude Desktop, create a new project.
3. Open the project's Instructions field and paste the full contents of Docs/Instructions.md.
4. Upload all seven reference files as project knowledge:
   • Docs/dynatrace_reference_guide.md
   • Docs/dynatrace_api_quick_reference.md
   • Docs/dynatrace_dql_reference.md
   • Docs/dynatrace_common_questions.md
   • Docs/dynatrace_troubleshooting_trees.md
   • Docs/dynatrace_bindplane_reference.md
   • Docs/dynatrace_bluebox_reference.md
5. Start a conversation in the project and ask a Dynatrace question.

Reference files

File	Covers	Staleness window
Docs/Instructions.md	Agent persona, routing rules, guardrails	Author-controlled
Docs/dynatrace_reference_guide.md	Core platform architecture: OneAgent, ActiveGate, Grail, deployment modes, OpenPipeline	~90 days
Docs/dynatrace_api_quick_reference.md	Classic vs. Platform API generations, auth models, token scopes, common endpoints	~90 days
Docs/dynatrace_dql_reference.md	DQL syntax, pipeline model, Smartscape topology queries, performance tuning	~90 days
Docs/dynatrace_common_questions.md	Pre-structured FAQ for pre-sales and support scenarios	~90 days
Docs/dynatrace_troubleshooting_trees.md	Decision trees for OneAgent, Kubernetes, API auth, ingest, and OpenPipeline issues	~90 days
Docs/dynatrace_bindplane_reference.md	Bindplane architecture, pipeline configuration, Dynatrace integration status	~30 days
Docs/dynatrace_bluebox_reference.md	Bluebox AI SRE agent: workspace setup, investigations, GitHub integration	~14 days

Staleness policy

Every reference file carries a Last verified date and a staleness window. The agent treats these as freshness signals:

• ~90 days (core Dynatrace files): re-verify before answering questions involving specific version numbers, UI menu paths, API field names, or token scope names.
• ~30 days (Bindplane): acquired April 2026; integration depth with core Dynatrace changes frequently. Re-verify integration claims every time.
• ~14 days (Bluebox): launched weeks before initial verification; CLI flags and setup steps change at high frequency. Treat every specific detail as a hypothesis to confirm against docs.bluebox.ai.

Live documentation always wins over the reference files. When a discrepancy is found, the agent flags it explicitly.

Contributing

See CONTRIBUTING.md for how to update reference files, report stale content, or propose new coverage.

License

This work is licensed under Creative Commons Attribution 4.0 International. See LICENSE for details.