DIMETRICS MCP

Every HPC lease term, every capital-stack tranche, every MW number: in your Excel cell.

Ask Claude or Codex any structured question about a publicly-disclosed HPC deal, an operator's power capacity and pipeline, the debt behind a lease, or the political risk on any site: and get a sourced answer in under 5 seconds. No more grepping 10-Ks for the escalator. No more reconstructing the debt schedule by hand. The same dataset, in every tool you already use.

$1,999 / year · or $199/mo · 5,000 tool calls/mo

What's in the dataset

Operators covered

33+

BTC-miner, neocloud (incl. pre-IPO) and hyperscaler operators

HPC deals tracked

35+

lease-economics with full terms

Source

SEC EDGAR

10-K, 10-Q, 8-K, S-1

Update cadence

T+1

trading day after EDGAR filing

Universe

MARARIOTCLSKCORZWULFCIFRHUTBTDRHIVEKEELGLXYAPLDSLNHVIPBGDEANYIONDCRWVNBISIRENDGXXSHAZWYFICHRNDUOTAMZNGOOGLMETAMSFTORCLSPCXNSCLCRUSO+ more added monthly

Every metric returned by the MCP carries a source_filing URL and as_of_date so you can reconcile back to the original 10-Q or 8-K.

What you can query: per company, across the full universe

  • Lease & deal economics: per-deal lease rates ($/kW/mo on critical-IT and gross), total contract value, annual NOI and margin, term length, tenant, PUE, and expected delivery dates.
  • Power capacity & pipeline: per-site energized and pipeline gross MW, critical-IT MW where disclosed, build status with provenance confidence tags, state / county / country, grid operator, utility, load zone, acreage, and estimated energization year.
  • Debt & financing: the instruments financing each lease: principal, coupon, maturity, instrument type, announcement date, with filing citations: verified links only, never inferred.
  • GPUaaS offtake economics: operator compute-sale agreements: total contract value, annualized revenue, and the implied $/GPU-hr rate (derived: a fleet-average-anchored floor, not a disclosed price; gpu_count usually estimated), plus which debt instruments finance them.
  • Political & regulatory risk: per-site political risk scores (0–100, raw and build-status-capped), risk bands, county moratorium exposure flags, and MW-weighted company rollups: the same methodology as the Dimetrics political-risk map.
  • Aggregations: every table slices by site, by company, or by energization year, and scalar metrics return as single values built to drop straight into a spreadsheet cell.

Quick start

You'll need an active subscription. Sign up or subscribe on your account page.

Set up Claude for Excel

  1. 1
    Open Excel and load the Claude add-in. The right-side pane shows the Claude chat interface.
  2. 2
    Click the gear icon in the Claude pane → ConnectorsAdd custom connector.
  3. 3
    Paste this URL as the MCP server URL (leave the OAuth fields blank: we'll register Claude automatically):
    https://www.dimetrics.ai/api/mcp
  4. 4
    Click Add. A browser tab opens to our consent screen showing the permissions Claude is requesting. Click Approve.
  5. 5
    You'll bounce back to Claude. The connector is now active: start a new chat and ask a question that needs deal data.

Set up Claude Code

  1. 1
    Go to your account pageMCP API keysGenerate new key. Give it a name (e.g. "Macbook · Claude Code") and copy the dmk_live_... value. It's shown once.
  2. 2
    In your terminal, register the connector:
    claude mcp add dimetrics --transport http --header "Authorization: Bearer dmk_live_YOUR_KEY" https://www.dimetrics.ai/api/mcp
  3. 3
    Reload your Claude Code session. Test with: claude mcp list: Dimetrics should appear.

Set up OpenAI Codex

  1. 1
    Add the server, then sign in: no API key needed:
    codex mcp add dimetrics --url https://www.dimetrics.ai/api/mcp codex mcp login dimetrics
    Your browser opens to our consent screen: sign in to Dimetrics if prompted, click Approve, and you'll bounce back to the terminal.
  2. 2
    Verify with codex mcp list: Dimetrics should show enabled with Auth OAuth. The Codex CLI, IDE extension, and desktop app all read the same ~/.codex/config.toml, so you only do this once.

Prefer an API key? (headless machines, CI)

Generate a key on your account pageMCP API keys, then add this to ~/.codex/config.toml:

[mcp_servers.dimetrics] url = "https://www.dimetrics.ai/api/mcp" http_headers = { "Authorization" = "Bearer dmk_live_YOUR_KEY" }

The --bearer-token-env-var flag also works if you'd rather keep the key out of the file: but the variable must be set in the environment that launches Codex. IDE extensions started from the Dock/Start menu don't read your shell profile, which is the most common reason that variant fails.

Set up Claude Desktop

  1. 1
    In Claude Desktop, go to SettingsConnectorsAdd custom connector.
  2. 2
    Paste the MCP URL:
    https://www.dimetrics.ai/api/mcp
  3. 3
    Click Add → approve on the consent screen → done. Same OAuth flow as Excel.

Note: if Desktop has trouble opening the consent screen in-app, paste the OAuth callback URL into your browser directly. Anthropic is rolling out improved consent handling for Desktop.

Set up OpenClaw (and other agent frameworks)

  1. 1
    Generate an API key on your account pageMCP API keys. Headless agents should use an API key: the OAuth flow needs an interactive browser consent that an autonomous agent can't complete.
  2. 2
    Register the connector (one command: it persists to ~/.openclaw/openclaw.json):
    openclaw mcp set dimetrics '{"url":"https://www.dimetrics.ai/api/mcp","transport":"streamable-http","headers":{"Authorization":"Bearer dmk_live_YOUR_KEY"}}'
  3. 3
    Verify the handshake: openclaw mcp doctor dimetrics --probe. All fifteen tools are auto-discovered: your agent picks the right one per question; discovery calls are free.
  4. 4
    Any other MCP-capable agent framework works the same way: point it at https://www.dimetrics.ai/api/mcp with the streamable-http transport and your key as a Bearer header.

Building an autonomous loop? Budget for the limits: 5,000 tool calls/month and 60 calls/minute per account: cache results and query incrementally rather than re-pulling the universe each cycle (429s are retryable with backoff). The Data License & Terms apply to what your agent does with the data, including anything it republishes.

Ready to wire it in?

Pro: $1,999/yr or $199/mo · 5,000 tool calls/mo

Sample prompts

Drop these into Claude (any surface). The MCP exposes seventeen tools: get_deal_metric, list_company_deals, get_available_capacity, get_debt_for_lease, get_political_risk, get_moratorium_summary, list_datacenters_at_risk, get_county_risk, get_capital_structure, get_instrument_detail, reconcile_entity_debt, list_gpuaas_deals, get_gpuaas_metric, get_gpuaas_kpi, get_debt_for_gpuaas, get_lease_terms, get_offtake_terms: and Claude picks the right one based on your question. Every response carries source-filing references so you can verify. Companies in the universe are queryable for sites and power pipeline even when they haven't signed a lease yet.

Lease rate, single deal

What's Cipher's crit-IT lease rate at the Black Pearl site?

Response

{
  "operator": "CIFR",
  "site": "Black Pearl",
  "tenant": "Amazon",
  "metric": "crit_it_lease_rate_per_kw_month",
  "unit": "$/kW/month",
  "value": 152.78,
  "source_filing": "CIFR 8-K 2025-11-14",
  "as_of": "2025-11-14"
}

Operator-wide aggregate

What is Cipher's weighted-average crit-IT lease rate across all disclosed deals?

Response

{
  "operator": "CIFR",
  "metric": "crit_it_lease_rate_per_kw_month",
  "unit": "$/kW/month",
  "value": 137.83,
  "agg": "avg",
  "n_deals": 3,
  "per_deal": [
    { "tenant": "Amazon", "site": "Black Pearl", "value": 152.78 },
    { "tenant": "Fluidstack", "site": "Barber Lake", "value": 130.42 },
    { "tenant": "Fluidstack", "site": "Reveille", "value": 130.30 }
  ]
}

Excel cell formula

=DIMETRICS("get_deal_metric", operator="WULF", metric="annual_contract_noi")

Response

147300000   (returns the numeric value, ready for downstream math)

Cross-operator comp

Rank CIFR, WULF, IREN, CORZ, and APLD by weighted-average $/kW/month lease rate. Highest first.

Response

Claude calls get_deal_metric 5x and returns:

  CIFR    $137.83/kW/month   (3 deals)
  IREN    $135.21/kW/month   (2 deals)
  APLD    $128.40/kW/month   (4 deals)
  WULF    $122.50/kW/month   (3 deals)
  CORZ    $114.66/kW/month   (2 deals)

Site pipeline lookup

What is the energization date and gross MW capacity for Cipher's Colchis site?

Response

CIFR: Colchis, TX  (pipeline)
   Gross capacity (pipeline)  1,000 MW
   Pipeline tier              load_study  (regulatory funnel: T-24mo)
   Target energization        2028
   Status                     Pipeline
   Source                     CIFR site disclosures (sites table)

Debt terms on a named lease

What are the debt terms on Cipher's Amazon lease at Black Pearl?

Response

CIFR: Black Pearl, Wink TX  (tenant: Amazon)
   Issuer            Black Pearl Compute, LLC  (CIFR subsidiary)
   Instrument        Senior secured notes
   Principal         $2.00 B
   Coupon            6.125% fixed
   Maturity          Feb 2031  (≈ 5-yr term)
   Use of proceeds   Black Pearl AWS DC construction
                     + reimburse CIFR equity contributions
   Source            8-K 2026-02-11

Delivery risk on a named lease

What happens if Cipher delivers the Black Pearl campus to Amazon late?

Response

get_lease_terms → CIFR-AMAZ-016 (tenant: AWS)

late_delivery_termination  [SUMMARY tier]
  "Amazon, as tenant, may terminate the lease if delivery is
   delayed by 120 days and Cipher fails to cure within a
   subsequent 30-day window; Amazon also holds termination
   rights for casualty, condemnation, a grid power outage
   exceeding 180 consecutive days, and anti-bribery
   violations, while Cipher has no right to terminate for
   convenience."
  params: trigger_delay_days 120 · cure_period_days 30

capex_spend_obligations  [SUMMARY tier]
  "Cipher's development-cost ceiling under the Amazon lease
   is $9.5 million per IT megawatt; Amazon covers
   construction cost overruns above that ceiling…"
  params: capex_ceiling_usd 9,500,000 per_mw

not_filed: sla_availability, events_of_default:
  searched, nothing public. Protections not visible ≠ no risk.
Source: CIFR 8-K EX-99 (EDGAR link on every row)

Credit support across an operator’s leases

Who backstops Cipher’s leases, and with what?

Response

get_lease_terms → guaranty_credit_support, all 3 CIFR leases

Black Pearl (AWS): "Amazon.com, Inc., the public parent,
  unconditionally guarantees all base rent and operating
  expenses owed by the Amazon tenant entity for the life of
  the lease… an AA-rated hyperscaler counterparty."

Barber Lake (Fluidstack): "Google agrees to backstop
  $1.4 billion of Fluidstack's lease obligations to support
  the project's debt financing, receiving in exchange
  warrants for roughly 24 million Cipher shares (about a
  5.4% pro forma stake)…"

Stingray (AWS): "Amazon.com, Inc. fully guarantees all base
  rent and operating expenses owed under the AWS lease…"

More questions to ask

A starting menu of what the dataset can field, grouped by domain. Each line is a prompt you can paste straight into Claude or Codex: click any one to copy it. The tags show which tools back each domain; you never name a tool, Claude picks. The last group spans tools: the kind of cross-referenced question that's slow to assemble by hand from filings.

HPC lease & deal economics

get_deal_metriclist_company_deals

Power portfolio & owned sites

get_available_capacity

Project financing (debt behind a lease)

get_debt_for_lease

Contract execution risk (lease & offtake terms)

get_lease_termsget_offtake_terms

Company capital structure

get_capital_structureget_instrument_detailreconcile_entity_debt

Datacenter geopolitical & regulatory risk

get_political_riskget_moratorium_summarylist_datacenters_at_riskget_county_risk

GPUaaS & neocloud offtake

list_gpuaas_dealsget_gpuaas_metricget_gpuaas_kpiget_debt_for_gpuaas

Cross-referencing (spans multiple tools)

get_deal_metricget_debt_for_leaseget_political_riskreconcile_entity_debt

Pricing

Pro: Monthly

$199/mo

5,000 tool calls/mo

The full toolkit: every dataset, every screen, and the MCP

  • Everything in Free, plus the full datacenter layer
  • Miner-to-HPC conversion dashboard: leases, MCap/MW, capacity bridge
  • GPU compute index: 38 providers, spot rates, forward curve, ClusterMAX
  • Colocation lease and GPUaaS offtake comps, every figure EDGAR-cited
  • Site-level capacity, debt-per-lease linkage, full capital structure
  • Moratorium and political-risk analysis with county PRS scoring
  • HPC lease calculator with real-deal pre-fill
  • Dimetrics MCP in Claude, Codex, and Excel: 5,000 calls/mo, OAuth + API keys

Pro: Annual

$1,999/yr

save 16% vs monthly · 5,000 tool calls/mo

The full toolkit: every dataset, every screen, and the MCP

  • Everything in Free, plus the full datacenter layer
  • Miner-to-HPC conversion dashboard: leases, MCap/MW, capacity bridge
  • GPU compute index: 38 providers, spot rates, forward curve, ClusterMAX
  • Colocation lease and GPUaaS offtake comps, every figure EDGAR-cited
  • Site-level capacity, debt-per-lease linkage, full capital structure
  • Moratorium and political-risk analysis with county PRS scoring
  • HPC lease calculator with real-deal pre-fill
  • Dimetrics MCP in Claude, Codex, and Excel: 5,000 calls/mo, OAuth + API keys

Subscriptions are governed by the Data License & Terms : individual use, no redistribution or resale.

FAQ

© 2026 Dimetrics · Data License & TermsAbout MCP