New: intraday auctions Β· wind & solar forecasts Β· the free live map
Prices, forecasts, weather stations, imbalance, flows and outages β 43 zones, 15-minute native, six in-house Volt models, one API, a free live map and a trading desk. And one promise nobody else makes: 17% less forecast error than the free baseline β or your zone is free. Scored in public, daily.
30 seconds Β· no credit card Β· free zone forever Β· paid plans from β¬9/mo
β17%
forecast error vs baseline
43
bidding zones
<60s
auction webhook
DE-LU Β· day-ahead
15-minute periods Β· EUR/MWh
Real day-ahead prices, last 30 days β not a hypothetical. Pick your zone:
Savings, last 30 days
β¦
Computing from realized pricesβ¦
β β¦/year
Start capturing this βWhy now Β· why Voltcast
Europe's grid got complicated fast β 15-minute markets, negative prices, weather-driven everything. The data to navigate it was split between hobby feeds and β¬3,500+/yr exchange licenses. We rebuilt the whole stack, then kept going.
2025-10-01 Β· the switch
The day went from 24 prices to 96.
Most integrations broke.
15-minute native everywhere, an explicit resolution_flag on every row, and the hourly-history seam handled β so it never surprises you. German Β§41a EnWG dynamic tariffs and 95% smart-meter rollout by 2030 make this the decade's data problem.
43
bidding zones Β· history to 2015 Β· bulk parquet
6
Volt models in production β prices, wind & solar, spreads, risk, imbalance, temperature
Scored in public. Losses included.
Daily verification vs free baselines β beat it or your zone is free, in the contract
43
weather stations β the exact gauges temperature markets & CME contracts settle on
The Voltcast Desk
A real-time trading desk with error bars on every tile β free demo, no login
Open the demo β
11
live map layers β prices, carbon, wind, temperature, uncertainty Β· free, no login
<60s
auction β your webhook, HMAC-signed Β· SSE streams Β· MCP for AI agents
125+
months of research + a new data-guide every day β every figure from our own database
Home Assistant to hedge desk
Python & TS SDKs, HA, evcc, Grafana, n8n, Sheets β from β¬9/mo to Quant
{
"zone": "DE-LU",
"model_version": "volt-2026.07.10",
"accuracy": "https://voltcast.com/accuracy/DE-LU",
"curve": [
{ "start": "2026-07-11T11:00:00Z",
"resolution_flag": "PT15M",
"p10": 41.20,
"p50": 58.75,
"p90": 84.10 },
...
]
}
Probabilistic forecasts
Per-zone gradient-boosted quantile models, retrained daily after the auction on load forecasts, wind & solar generation forecasts, weather and calendar structure. Horizons from 1 hour to 7 days, evaluated strictly walk-forward β no lookahead, ever.
The engine
Every forecast on this platform β prices, wind & solar, intraday spreads, imbalance, negative-price risk β comes from one in-house model family, trained on our own recorders and scored in public every single day. No vendored black boxes. No borrowed numbers. Ours.
VOLT 1
calendar + lags
VOLT 2
+ weather
VOLT 3
+ grid drivers
VOLT 4
+ TSO grid forecasts
VOLT 5
stacked ensemble + conformal
A generation ships only when it beats the one before it on the public scoreboard β walk-forward, no lookahead, losses included.
VOLT 4
P10/P50/P90 day-ahead curves, 1hβ7d, every zone. The flagship β up to a third less error than the naive baseline where it serves.
Volt 5 successor in training β
VOLT RES 1
Our own generation forecasts, scored head-to-head against the TSO's official forecast in every API response. Wins and losses both published.
vs the TSO, daily β
VOLT SPREAD 1
Predicts tomorrow's IDA1βDA spread per 15-minute slot, issued before the intraday gate closes β forward-blind by construction.
pre-IDA1 gate β
VOLT RISK 1
P(price < 0) per period, 14 days out, Brier-scored daily against climatology. The tracker's forward view.
Brier-scored β
VOLT TEMP 1
Probabilistic daily Tmax/Tmin on the exact stations markets and CME weather contracts settle on β full distributions, 1Β°C buckets, verified vs METAR daily.
station-exact, CRPS-scored β
The whole family, one API call away. Full quantiles, wind & solar and optimization start at β¬199/mo β and if Volt doesn't beat the free baseline in your zone, that zone costs you nothing.
New Β· Quant tier
A real-time decision workspace fusing auctions, intraday, imbalance, outages, weather stations and every Volt model β with the data's age and the models' verified error bars on every tile. Event tape with measured publication latencies. Run drift showing what changed overnight. Four persona presets, βK command palette, your layout saved.
Demo shows the full desk on yesterday's data β no login needed.
The scorecard
Pinball loss and MAE per zone and horizon, versus naive persistence and public free models β recomputed every day, in the open. You will never find a static accuracy claim on this site; the live page is the claim.
IE-SEM Β· day-ahead
β33.4%
error vs free baseline
MAE 26.76 vs 40.2 β¬/MWh
view the full record β
DE-LU Β· day-ahead
β31.0%
error vs free baseline
MAE 23.91 vs 34.64 β¬/MWh
view the full record β
EE Β· day-ahead
β27.8%
error vs free baseline
MAE 35.24 vs 48.8 β¬/MWh
view the full record β
Live numbers from the scorecard β walk-forward backtest until live evaluations accumulate, recomputed continuously.
The standing offer
If our forecast doesn't beat the free baseline in your zone, you don't pay for it. Zone-level credits, automatic, no support ticket required.
Developer-first
# 15-minute day-ahead prices, DE-LU, native resolution
curl -H "Authorization: Bearer $VOLTCAST_KEY" \
"https://api.voltcast.com/v1/prices/DE-LU?from=2026-07-10&to=2026-07-12"
import voltcast
vc = voltcast.Client("vk_live_β¦")
df = vc.prices("DE-LU", start="2026-07-10", end="2026-07-12").to_pandas()
import { Voltcast } from "@voltcast/sdk";
const vc = new Voltcast("vk_live_β¦");
const curve = await vc.prices("DE-LU", { from: "2026-07-10", to: "2026-07-12" });
Webhooks: auction_published, price_below, negative_prices β HMAC-signed, exponential retries
Optimization: cheapest-window & EV/battery charge schedules straight from the forecast curve
Bulk export: parquet/CSV, signed URLs, zero egress fees β your models, your data
Ecosystem: Python & TypeScript SDKs, Home Assistant component, evcc tariff provider
Grid & carbon: load, generation mix and carbon intensity + green score per 15-minute period β charge when it's greenest, not just cheapest
AI agents: native MCP server β plug Voltcast into Claude, Cursor or your own agents in one config block
Batteries & flex: FCR/aFRR/mFRR reserve prices, outage alerts, and a battery arbitrage backtest against real history
Everywhere you work: Grafana, n8n, Google Sheets, a free embeddable price widget, and monthly market reports
Pricing
β¬0
non-commercial Β· no card
β¬9/mo
Home Assistant Β· evcc Β· hobby
β¬49/mo
commercial use
β¬199/mo
99.9% SLA
β¬499/mo
invoice Β· DPA Β· procurement
The full balancing dataset for BESS & flexibility desks: reBAP + 6 more imbalance markets (revision-aware), FCR/aFRR/mFRR reserve prices, outages, imbalance forecasts, wind/solar forecasts + weather, zone temperature, the ML intraday-spread forecast (volt-spread-1), and bulk CSV/Parquet export.
Station-exact temperature for market & hedging desks: probabilistic daily Tmax/Tmin in 1Β°C buckets on the exact gauges markets settle on (Paris = Le Bourget), CME-city degree-day trackers with settle projections, population-weighted zone temperature, and station-integrity monitoring. Verified daily, losses included.
For teams training their own models: the Voltcast Desk (real-time multi-panel workspace), point-in-time-correct ML feature matrices β the exact training data our forecasts eat, anti-lookahead semantics documented per column β full-history training parquet exports, station temperature matrices, the as-of reconstruction API, and everything in Scale.
Multi-seat, SLA beyond Scale, battery revenue underwriting, custom zones or redistribution licensing β talk to a human, get an answer the same day.
Free key, no card. Every plan is self-serve β subscribe, upgrade or cancel from the dashboard. And every forecast we sell is scored publicly, every day, on /accuracy.
Secure payment via Stripe Β· annual β15% Β· cancel any time