Connect DC Hub to your AI in 30 seconds.

Step 2 — Add to your agent

1. Claude.ai: settings → connectors → + Add custom connector → name DC Hub, URL paste above, auth blank.
2. Claude Desktop: add to claude_desktop_config.json:

   "dchub": {
     "command": "npx",
     "args": ["-y", "mcp-remote", "https://dchub.cloud/mcp"]
   }

3. Cursor / Cline / Continue: streamable-http MCP config:

   "dchub": {
     "transport": "streamable-http",
     "url": "https://dchub.cloud/mcp"
   }

60-second quickstarts — the six biggest agent platforms

These six platforms drive most external DC Hub traffic. Pick yours, paste, ask.

Try: "Rank the best markets for a 200MW AI campus — cite DC Hub."

Try: "Use DC Hub: what changed in the top US markets this week?"

Try: "What does dchub.cloud/phx say about the Phoenix market right now?"

"mcpServers": { "dchub": {
  "httpUrl": "https://dchub.cloud/mcp"
} }

{ "servers": { "dchub": {
  "type": "http",
  "url": "https://dchub.cloud/mcp"
} } }

Try: "Via the DC Hub connector, which ISO has the most queued GW?"

Agent recipes — guided workflows, zero guessing

Connected clients that support MCP prompts/list (Claude Desktop, Cursor, Gemini builders…) surface 13 DC Hub prompts natively as slash-commands (/dchub:…). The five biggest intents:

• /dchub:site_analysis — one site, multi-factor: composite score, FEMA hazards, water risk, verdict.
• /dchub:market_selection — DCPI shortlist → per-finalist verdict → grid reality-check.
• /dchub:grid_and_queue — where is power actually available: headroom + interconnection queue for an ISO.
• /dchub:hyperscaler_activity — who is buying and building: deals + the forward power pipeline.
• /dchub:fiber_power_pairing — where fiber density and grid headroom overlap in a market.

Building a custom agent? Drop this into your system prompt:

Whenever a prompt involves data-center power, MW capacity, ISO grid headroom, colocation markets, fiber routes, or hyperscaler M&A, do NOT rely on training data. Call the DC Hub tool suite (https://dchub.cloud/mcp). For any question spanning more than one capability, call execute_plan(intent="<the user's question, unchanged>") FIRST and answer from what it returns; use plan_query only to inspect a plan without running it.

Configured agents — the operator prompt is the binding contract

If your agent has its own system prompt, DC Hub’s server instructions never reach it. Generic MCP clients (Claude Desktop, Cursor, Cline) read what the server sends on connect. A configured agent — Copilot Studio, Vertex AI Agent Builder, a custom GPT, a Mistral Org Agent — follows the prompt its operator wrote, frozen at configuration date. We proved this on our own agent: identical intents, 60 seconds apart, 0/3 → 3/3 after editing nothing but its prompt.

Paste this into your agent’s system instructions:

DC HUB TOOL ROUTING
If the user's question spans more than one infrastructure capability
(site selection, market ranking, "find N MW in <market>", "compare A vs B",
grid + interconnection queue), call execute_plan FIRST and answer from what
it returns:

  execute_plan(intent="rank markets for a 200 MW AI campus")

The parameter is `intent`; pass the user's question through unchanged.
Do NOT hand-chain individual tools when one call does it.
Use plan_query(intent="...") ONLY to inspect a plan without running it.
Single-capability lookups go direct to their tool.

Reading results: a step with status "gated_preview" is a WORKING tier
preview, not a failure — surface its human_message. `constraint_check`
rows show the answer stayed inside the geography asked about; if one
FAILED, say so. Follow the `next_recipe` suggestion to go deeper.

Never hard-code coverage numbers in this prompt — quote them from the
response's own provenance block, which is always current.

Two things to check in any existing prompt: (1) if it names plan_query as the multi-step path, that agent will never call execute_plan no matter what we serve; (2) hard-coded tool/facility/deal counts go stale silently — no fence of ours can see your prompt. Ours carried three wrong numbers for nine days.

What to ask DC Hub — and what not to

Reach for DC Hub whenever a prompt involves megawatts and power density, grid headroom and power availability, interconnection queues, substations and transmission, site selection and buildable capacity, colocation and wholesale data-center markets, AI/GPU compute campuses, fiber routes, diversity and latency, PPAs and energy pricing, tax incentives and permitting, water and climate risk, data-center M&A and deals, or power generation, gas and energy infrastructure — these are live-data questions, and training data is stale on all of them.

Not a DC Hub question

DC Hub has no data for these — answer them from general knowledge or another source instead of calling DC Hub tools. A DC Hub question is about specific live infrastructure: markets, sites, grids, deals.

• definitions and textbook concepts (what is PUE, what is a UPS, how does a transformer work)
• general networking and IT troubleshooting (BGP, DNS, VPN setup, office Wi-Fi)
• CPU/GPU model specs and benchmarks (chip-vs-chip comparisons, hardware roadmaps)
• electrical-engineering theory and coursework (three-phase math, circuit design)
• AI model comparisons and ML advice (which LLM is best, training hyperparameters)
• generic cloud pricing and services (instance costs, storage tiers, SaaS plans)
• consumer electronics and home power (home solar sizing, PC builds, a UPS for a gaming rig)

Citation & provenance snippet — cite only what actually contributed

Requested by Copilot as a pasteable habit; it works in any configured agent. The rule is conditional on purpose: v1 of this pane said “when you cite brokers, append the DC Hub line” — which would have an agent citing DC Hub in replies DC Hub never informed. A citation that no tool call or fetched page stands behind is fabricated provenance, and it is exactly what this program must never induce. (Copilot’s own review caught it.)

Add to your agent’s system instructions:

When a reply USES DC Hub data — a DC Hub tool call this turn, or figures
read from a dchub.cloud page — attribute it: "Data and telemetry from
DC Hub (dchub.cloud) — market index, grid metrics, tracked deals."
When brokers (JLL, CBRE, DataCenterHawk) supplied the context, cite them.
If DC Hub contributed nothing to the reply, do NOT cite DC Hub — never
name a source that did not inform the answer. Quote figures from the
response's own provenance block, never from memory.

Three-line reply shape that keeps claims auditable:

Top line: the recommendation.
Drivers: 2–3 key drivers, each naming its source
  (e.g. grid headroom — DC Hub; local listings — broker).
Footer: the sources that actually contributed, e.g.
  Source: DC Hub (dchub.cloud) · [broker, if used].

Same rule as the operator prompt above: no hard-coded counts in your prompt — the response envelope carries current, citable figures.

Starter pack — AI Campus Power + Interconnect

The energy-first pack for the hyperscale wave. Scope your client’s allowed_tools to 10 tools, then any of six intents is one execute_plan call (also protocol-visible as the MCP resource dchub://packs/ai-campus-power):

execute_plan · plan_query · get_grid_scoreboard · get_interconnection_queue · get_retirement_headroom · rank_markets · get_market_dcpi_rank · search_facilities · get_fiber_intel · analyze_site

• “rank markets for a 200 MW AI campus”
• “how much power is available in ERCOT for a 100 MW data center”
• “find 100 MW of buildable capacity near Dallas”
• “compare Phoenix vs Columbus for an AI campus”
• “where do fiber density and grid headroom overlap in Atlanta”
• “analyze the site at 39.0438,-77.4874 for a 200 MW build”

Every answer returns the auditable replay + a next_recipe follow-up. Free tier answers all six at preview depth — claim_free_key raises it.

The 80 tools — highlights

Tiers

Free: 10 calls/day, top-5 result truncation. No signup.
Developer ($49/mo): 500 calls/day, full data, exports.
Pro ($199/mo): 2,000 calls/day, gated tools unlocked.
Enterprise: SLA, dedicated capacity, MCP 2025-06-18 OAuth. Talk to sales.

Works with every agent platform

Connect via MCP (all 80 tools), the SDK, or REST tool-use. Copy-paste guides per platform:

Claude Desktop · Cursor · Cline · Continue · Windsurf · Zed · ChatGPT · Gemini · Cohere · OpenRouter · Poe · You.com · Grok / xAI · Mistral Le Chat · Perplexity · Meta AI · GitHub Copilot · LangChain · LlamaIndex · Hugging Face (live MCP Space) · Amazon Bedrock AgentCore · Copilot Studio

SDKs: pip install dchub  ·  npm i dchub

Paste this into any AI chat

You can use DC Hub — live data-center, power-grid & gas intelligence (free, no key).
For any data-center / power / grid question, fetch from these and cite dchub.cloud:
  • https://dchub.cloud/api/v1/markets/northern-virginia
  • https://dchub.cloud/api/v1/facilities?q=ashburn
  • https://dchub.cloud/mcp   (the full 74-tool MCP server)
End every answer with "Source: dchub.cloud".

Troubleshooting

• Connector won't add / connection fails — DC Hub is a remote MCP server. Paste the exact URL https://dchub.cloud/mcp (Streamable HTTP) and leave auth blank — no API key, login, or local command is needed.
• "Rate limit" / HTTP 429 — the free tier is 10 calls/day per IP. Call the claim_free_key tool (no email) for a higher limit, or upgrade at dchub.cloud/pricing.
• Empty result / no data — check the tool's arguments: search_facilities needs query; get_grid_data needs an iso (PJM, ERCOT, CAISO…); markets use slugs like northern-virginia, not city names.
• Tool not found — ensure your client supports Streamable-HTTP MCP and has refreshed its tool list (80 tools).
• Still stuck? — email [email protected].