DOE-MCP

Case study · Policy

Five read-only servers for public energy, research, earth, materials, and biology data. Independent project; answers include sources and coverage limits.

Stack: Python; five read-only MCP servers and a source registry.
Status: Early · five servers work against live endpoints
Dates: Reviewed November 6, 2026; data dates shown in the app
Links: Live site · Source code

Background

How can an assistant query public DOE data and identify the limits of an answer?

DOE and national-laboratory data span unrelated research, operations, earth-science and materials systems. DOE-MCP divides that work among five read-only servers and describes both active sources and inactive inventory.

This is an independent project, unaffiliated with or endorsed by DOE, a national laboratory or a program office. Registry totals describe that published inventory rather than every DOE dataset.

Screenshots

Select a screenshot to enlarge it.

Five focused MCP servers expose public DOE and national-laboratory research (mobile) — Five focused MCP servers expose public DOE and national-laboratory research, energy, earth, materials, and biology data with provenance and coverage on every answer.

Five focused MCP servers expose public DOE and national-laboratory research (desktop) — Five focused MCP servers expose public DOE and national-laboratory research, energy, earth, materials, and biology data with provenance and coverage on every answer.

The generated source registry shows what each public system holds (mobile) — The generated source registry shows what each public system holds, who stewards and funds it, how many records it exposes, and why an inventoried source is not active yet.

The generated source registry shows what each public system holds (desktop) — The generated source registry shows what each public system holds, who stewards and funds it, how many records it exposes, and why an inventoried source is not active yet.

Features

Focused servers: Research, energy, earth and materials expose typed queries through a shared response contract.
Source registry: Manifests identify publishers, access requirements and activation gaps.
Evidence envelope: Answers separate data, provenance, evidence, coverage, warnings and execution details.
Empty results: A covered query with no match remains distinct from a missing or inactive source.
Generated reference: Assembled tools and registry entries generate the site and tool documentation, with drift checks in CI.

Reviewed November 6, 2026 · Project documentation.

[View the Live Site] | [View the Code]

← Back to Homepage