InfraSure

Search US power plants

Search the InfraSure US power-plant directory (15,000+ operating plants) by free-text query, fuel type, state, ISO/RTO, capacity range, and owner. Returns slim entity rows — each result includes plant_id (pass to `get_plant` for full detail) and slug (URL fragment: InfraSure/plant/<slug>). Use this tool to discover plants matching a description; use `get_plant` to fetch comprehensive data on one.

Get full plant data

Fetch comprehensive public detail for one plant by EIA plant_id. Returns ownership, generators, grid attributes, generation history, financials (when available), boundary geometry, nearby plants, similar plants, resolution-layer news_extracted facts, and an additive pricing_context. Annual pricing is attached only through accepted generator-grain links in the current published pricing release; InfraSure never infers a plant node from proximity or a legacy grid label. Use after `search_plants` re

Get full interconnection-queue project data

Fetch the comprehensive JSONB blob for one queue project by slug (from search_projects results). Returns developer, technology, status, capacity, interconnection point, milestones, location, EIA-plant match (if any), pricing context, and resolution-layer news_extracted facts. Use after `search_projects`.

Search interconnection-queue projects

Search the InfraSure US interconnection-queue project directory (~9,800 projects: solar, wind, battery, gas, etc. in development across LBNL + 7 ISOs). Filters by free-text query, state, region (ISO/RTO), technology, status, capacity, developer, queue year. Returns slim project rows; pass slug to `get_project` for full detail.

Search US data centers

Search InfraSure's active production data-center directory by place or operator name, state, ISO/RTO, service model, and lifecycle status. min_mw/max_mw remain available only for the legacy flat release; canonical releases reject them until callers can declare an exact power-comparison cohort. When the canonical model is current, each result is one reviewed physical browse root—not a subsystem—with explicit campus/building/space type, nested-member counts, representative-point method, and select

Get public data-center detail

Fetch one production data-center place by canonical slug/ID or any active/redirect legacy slug/ID. When the canonical model is current, the response includes the reviewed browse-root type, representative-point method, separate physical children, operating presences and development objects, and selected power with its metric basis, reported grain, and scenario. A noncanonical handle resolves only through the release-pinned compatibility map and adds a public handle-resolution warning. Separately

Get annual pricing-node history

Fetch the current published annual wholesale-price history for one stable pricing_node_id returned by a plant or data-center pricing_context. Preserves market run and price type (LMP, LBMP, or settlement-point price), components when supported, source-backed versus modeled evidence class, fill method, source release, aggregation method, and observation coverage. Source-backed history is the default; modeled temporal backfills require include_modeled=true. This is historical market-node context—n

Search energy news

Search InfraSure's durable, editorial energy stories; operational weather alerts are always excluded. Filter by query, category/subcategory, plant, queue project, owner, entity mention, fuel, state, lane, event_type, and date range. Use category='hazards' with a subcategory such as 'hail' for realized infrastructure-impact stories, not warnings or modeled peril exposure. Entity filters match every linked row while results return one canonical primary article per story, newest first.

Get resolution-layer canonical facts

Returns the resolved facts (developer, EPC contractor, PPA terms, deal value, milestones, expected COD, etc.) extracted from news articles for a single plant or project. Each fact includes value, source attribution, as_of date, confidence, and how many articles supported it. STABLE facts use most-cited wins; TEMPORAL facts (milestones, status) use most-recent wins. This is InfraSure's unique enrichment layer — facts are pre-resolved, not raw article text.

Reverse-lookup entities by an extracted fact

THE DIFFERENTIATOR. Find every plant or project where an extracted fact has a specific value. Examples: every project where developer = 'primergy solar'; every plant where transaction_type = 'asset acquisition'; every project where action_type = 'lawsuit'. Supports stable + temporal fact fields. Returns matching entity_ids paired with the supporting fact metadata (source, as_of, confidence). No other public energy data source exposes this kind of resolved-fact reverse lookup.

Find InfraSure analysis methodologies

Rank compact InfraSure Insights methodology candidates for a decision question by domain, analysis family, driver, and actor. Returns candidates only; it never silently loads or runs one. By default only active/eval-passed methods are returned. Set include_deprecated=true to inspect draft/deprecated packages, and preserve their status as a limitation. Use get_methodology only after choosing a slug.

Load one InfraSure analysis methodology

Load the selected InfraSure Insights methodology by canonical slug. Without file, returns the loadable SKILL.md method plus a bundled-reference manifest, maturity, required inputs, blocked claims, and confidence rules. With file, returns only the exact manifest path requested. Never treat status='draft' as validated or use a bundled reference as live-state evidence.

Search governed ISO and federal live-state evidence

Search InfraSure's curated, dated primary-source evidence registry for current ISO/RTO and federal market state. Every result includes the operator or government source, as-of and review dates, decision relevance, allowed claims, and claims the source cannot prove. Use this for current queue-process, capacity, reliability, large-load, market-design, ancillary-service, transmission-planning, and regulatory questions. It is evidence, not methodology; preserve its blocked-claim envelope and never g

Compare 2-8 plants/projects side-by-side

Fetch a side-by-side grid of 2-8 entities (mixed plants and projects). Pass plant_ids as numbers and project slugs as strings in `entity_ids`. Returns one row per entity with the requested `dimensions` (or a default identification + scale + ownership/developer set). Use this when the LLM needs to compare attributes across entities in one call rather than several `get_*` calls.

Aggregate plants or projects (count / sum / avg) by dimension

Group rows of plant_index or project_index by a dimension and aggregate. entity = 'plants' or 'projects'; group_by = state | fuel | iso | technology | status | owner (not all dimensions apply to both — see schema). metric = count | total_capacity | avg_capacity. Optional filter scopes the WHERE clause before grouping. Examples: total solar capacity by state ({entity:'plants', group_by:'state', metric:'total_capacity', filter:{fuel:'SUN'}}); count of active projects by technology ({entity:'projec

Plants near a given plant

Returns the pre-computed list of nearby plants for a given plant_id (by Haversine distance over their EIA-reported lat/lon). Each result includes plant_id, name, primary_fuel, capacity, and distance_km. Optional fuel filter and radius_km cap. Use to find competitors, neighbors, or geographically-clustered fleets.

Plants owned by a specific company

List plants owned/operated by a company (case-insensitive substring match across canonical_owner, utility_eia_name, gem_parent_name). Optional fuel + state filters. Examples: NextEra Energy (525 plants), Duke Energy, Vistra. Prefer this over search_plants when the LLM specifically wants 'plants owned by X' — the focused naming improves selection.

Pre-computed news distribution statistics

Returns pre-computed counts across the news_items archive: by category, state, fuel, source, owner, data_source, plus the top 50 mentioned entities. group_by='all' returns the entire stats blob. Updated whenever news is pushed. Use for a quick overview of the news landscape without scanning individual articles.