ERCOT · Pricing
ERCOT real-time LMP
Locational marginal prices from the latest SCED run, every ERCOT settlement point. Collected from ERCOT MIS, reconciled against the source file.
ercot-realtime-lmpevery 5 min · 5m rows1,118 entitiesWhat is in it
- Updates
- every 5 min · 5m rows
- Entities
- 1,118 entities — HB_HOUSTON, HB_NORTH, LZ_WEST…
- Source
- ERCOT MIS · NP6-788-CD
- Source timezone
- US Central — every timestamp Dryos serves is UTC
- Coverage
- 15 August 2026 to now
- Rows held
- 8,719,267
- Primary key
- iso, market, node, interval_start_utc
- On the map
- 698 of 1,118 placed — EIA-860M plant coordinates, through ERCOT's unit mapping
What it measures
Every value, its unit, and what the source means by it.
- Total LMP$/MWh
The settled price. The only price field ERCOT populates.
lmp_total
Columns
Exactly what a request answers with, read from the API rather than written here.
| Column | Type | Description |
|---|---|---|
| interval_start_utcrequired | timestamp | Start of the settlement interval, UTC. Always UTC — the ISOs publish in local time with inconsistent DST handling.e.g. 2026-08-14T15:00:00Z |
| isorequired | string | Publishing ISO.e.g. ERCOT |
| marketrequired | string | Market the price cleared in: DAM (day-ahead), RTM (real-time, SCED), or RTPD for CAISO's fifteen-minute market.e.g. RTM |
| noderequired | string | Pricing node or settlement point identifier, as published by the ISO.e.g. HB_HOUSTON |
| node_type | string | Node classification. For ERCOT this is derived from the settlement point naming convention: HUB / LOAD_ZONE / DC_TIE / RESOURCE_NODE. For MISO it is the ISO's own pricing-node type: HUB (every aggregate, the eight trading hubs and the .AZ zones alike) / LOAD_ZONE / RESOURCE_NODE (a generator's node) / INTERFACE. For PJM it is the type on the row: HUB / LOAD_ZONE (a transmission zone) / INTERFACE / EHV / AGGREGATE for the aggregate streams, and RESOURCE_NODE (a generator's bus) / LOAD_BUS / EXTERNAL_BUS for the bus streams. For SPP it is read off SPP's own suffix on the Pnode behind the settlement location: HUB / LOAD_ZONE (a load area) / RESOURCE_NODE / DEMAND_RESPONSE / INTERFACE (a neighbouring operator). For CAISO it is the APnode type in CAISO's own reference: HUB (trading hubs) / LOAD_ZONE (every kind of load aggregation point) / RESOURCE_NODE / DEMAND_RESPONSE (a proxy or reliability demand resource, read off the resource id) / INTERFACE (intertie resources) / EIM_NODE (a Western EIM area's node — CAISO's EIMT, which holds the area's plants and its transfer points alike) / AGGREGATE (generation aggregation points, distributed-resource aggregations, flexible-ramp and imbalance-reserve zones). For NYISO it is LOAD_ZONE (the eleven zones) / INTERFACE (the four external proxy zones, the proxy buses of scheduled lines, and any generator NYISO files in an external zone) / DEMAND_RESPONSE (a demand-response provider) / RESOURCE_NODE (a bus in NYISO's generator reference, DER aggregations included) / LOAD_BUS (a bus NYISO prices beside the generators that no generator is registered to — transformer banks and load buses). For ISO-NE it is the location type ISO-NE publishes, a network node refined by the prefix on its name: HUB (the Internal Hub) / LOAD_ZONE (the eight) / INTERFACE (an external node) / DEMAND_RESPONSE (a demand-response aggregation zone) / HUB_NODE (one of the network nodes the Hub averages) / RESOURCE_NODE (a unit, UN., or a storage plant's asset-related demand, AR.) / LOAD_BUS (a load node, LD.).e.g. APND |
| lmp_totalrequired | number | Total locational marginal price, $/MWh.e.g. 41.83 |
| lmp_energy | number | Energy component, $/MWh.e.g. 39.12 |
| lmp_congestion | number | Congestion component, $/MWh. Signed so that total = energy + congestion + loss on every row: NYISO publishes it with the opposite sign, and it is stored negated.e.g. 2.94 |
| lmp_loss | number | Loss component, $/MWh.e.g. -0.23 |
| source_published_at_utc | timestamp | When the ISO published this interval, UTC. Lets a buyer distinguish 'we were slow' from 'the ISO was slow'. |
| collected_at_utcrequired | timestamp | When Dryos collected this row, UTC. Stamped by the platform, not the collector. Together with interval_start_utc this makes freshness auditable per row instead of only in aggregate.e.g. 2026-08-15T21:22:09Z |
Read it
Public, read-only, no key and no sign-up. Relative times stay relative, so a URL keeps working.
The newest rows
curl "https://api.dryos.ai/v1/datasets/ercot-realtime-lmp/query?node=HB_HOUSTON&limit=12"The catalogue entry — every column, its type and the source
curl "https://api.dryos.ai/v1/datasets/ercot-realtime-lmp"Hourly averages for the last day
curl "https://api.dryos.ai/v1/datasets/ercot-realtime-lmp/query?node=HB_HOUSTON&start=-24h&interval=1h&agg=avg"Full parameters — bucketing, aggregation, filters and search — are on the API reference.
Ask an agent
The same stream through the Model Context Protocol, so Claude, ChatGPT, Cursor or any MCP client can query it directly.
claude mcp add --transport http dryos https://api.dryos.ai/mcpThen ask:
What has Total LMP been at HB_HOUSTON over the last 24 hours, and when did it peak?
Setup for every client is on the MCP server page.
Changes to this collector
A change in the collector is a change in the data. Anyone reading a step in a series is owed the day it moved under them.
- 2026-09-08
Collects within the minute again; the record had read five to six minutes late on every interval.
The delivery record showed every interval collected five to six minutes after ERCOT posted it. Two causes. Each run re-fetched the previous SCED file — the document filter compared publish times against the interval watermark, which the newest file always passes — and the upsert overwrote collected_at with the second fetch, so the record kept the later time. And an off-grid SCED run shifted the predicted window by half a cycle, so the two regular files after it were each found two minutes late. The collector sits on the shared MIS base now, filtering on the publish watermark; the window is predicted from the median phase of recent publications; a 30-second heartbeat outside the window catches anything the prediction misses; and the window polls every 4s from 15s before the stamp, since the listing shows a file about 40s after the time it carries.
How this stream is checked
- Pulled straight from ERCOT MIS · NP6-788-CD — no reseller in between — and compared against that file before it is served.
- Null means the source did not report it — never zero.
- A reading lands every 5 min; the interval timestamp is the start of the interval, in UTC.
- ERCOT settles and publishes on US Central time. Every timestamp Dryos serves is UTC.
Related streams
- ERCOT day-ahead hourly LMPERCOT
Hourly cleared prices from the day-ahead market for every ERCOT electrical bus, posted once for the following day.
daily, ~12:35 CT · 1h rows19,312 entitiesercot-dam-lmp-bus - Real-time settlement point pricesERCOT
The 15-minute price settlement actually uses, every ERCOT settlement point — the SCED LMP plus price adders.
every 15 min · 15m rows1,118 entitiesercot-rt-spp - DAM settlement point pricesERCOT
Day-ahead hourly settlement point prices for every settlement point, posted once after the DAM run.
daily, ~12:35 CT · 1h rows1,118 entitiesercot-dam-spp - Real-time LMPs by electrical busERCOT
Bus-level prices under the settlement points: ~19,000 electrical buses from every SCED run.
every 5 min · 5m rows19,312 entitiesercot-rt-lmp-bus - Indicative LMPs (RTD look-ahead)ERCOT
Where real-time prices are about to go: RTD's forward intervals for every settlement point, republished each run with every vintage kept.
every 5 min, look-ahead · 5m rows1,118 entitiesercot-indicative-lmp - SCED system lambdaERCOT
The system-wide marginal energy price from every SCED run — one number for what energy is worth in ERCOT right now.
every 5 min · 5m rows1 entityercot-sced-lambda