add_project_file
ChatGPTStore a document and queue it to be read; one with the same name in the folder becomes its next version. Send text files as text, others as base64 content. Then poll get_project_file until extraction is ready. A planning case (PSS/E .RAW) is studied at the project's POI once read; an ISO contingency list (CSV) is used by every study after.
archive_alert
ChatGPTDeprecated compatibility alias. Archive and disable one owned alert only after the user explicitly asks to archive, remove, delete, or cancel it. Fresh clients should use manage_alert with action=archive.
create_alert
ChatGPTCreate one alert only after an explicit user request. Resolve exact targets first and copy the returned id unchanged; names such as LZ_AEN can resolve to node:LZ_AEN and must not be rewritten as zone targets. For numeric rules, compare the threshold with recent data and confirm values that appear almost always true, almost never true, or backwards for the stated goal. Alerts notify on state transitions, never per-second cadence.
create_project
ChatGPTCreate a private project for a question, e.g. whether a site can connect. Only when the user asks for a new project.
describe_data
ChatGPTInspect one exact dataset from search_data. Returns its full schema, keys, time semantics, and query contract inside the given market scope.
get_alert
ChatGPTRead one owned alert and its recent evidence, including an archived alert returned by list_alerts.
get_alert_capabilities
ChatGPTRead executable alert combinations, narrowed by the market and any known target, trigger, metric, or condition. Call before create_alert or an update when support is uncertain.
get_asset_hydration_status
ChatGPTReturn the durable workflow status for one hydration request owned by the authenticated user and this connection.
get_asset_intelligence
ChatGPTEnterprise: return the complete reviewed, source-backed dossier for one exact asset, including technical, market, financing, timeline, citations, and evidence boundaries.
get_asset_profile
ChatGPTReturn a compact, reviewed public profile for one exact asset id from query_grid_assets. If the asset exists without a published profile, the result says so explicitly.
get_congestion_path
ChatGPTMeasure one source→sink congestion path in one market: the constraints that move its basis, this week's outlook, and a hedge / instrument / ignore call at a stated MW size. Use exact settlement point ids from query_grid_assets.
get_live_grid_status
ChatGPTReturn current headline values and source-freshness metadata for one market.
get_project_file
ChatGPTWhether a document has been read (extraction: pending, ready or failed with readError), and for a planning case the study kept at each version.
get_saved_findings
ChatGPTGrid Radar findings and grid events reaching the user's saved assets, or one portfolio's, each with the assets it reaches and how: direct (names the asset or its settlement point), priced (a constraint's $/MWh effect at the asset's own node, by shift factor), located (hub, zone, county or distance), technology, or market-wide. Includes findings on the assets' own settlement point prices. Pass asOf to replay a past instant within 30 days.
get_wire_briefings
ChatGPTRead the operator briefings written over the wire: what changed in a closed time window, why it matters, and what to watch, with citations back to the publishers. Each briefing states its knowledge cutoff — it saw only stories collected before that instant. A briefing can report that nothing material happened; that is an answer, not an empty result.
list_alerts
ChatGPTList your alert rules in one organization (Personal unless organizationId names another) by lifecycle status. Use this to resolve an alert id before reading or changing it.
list_grants
ChatGPTList every active OAuth grant on your Ask the Grid account, across all connected MCP clients and resources (per-market and global).
list_project_files
ChatGPTEach document's id, kind, role, version and read state (extraction).
list_project_findings
ChatGPTFindings with their status, the documents they cite and the requirement they bear on.
list_projects
ChatGPTList the projects you can open in one organization (Personal unless organizationId names another), with ids for the other project tools and a link for a person.
list_saved
ChatGPTList the authenticated user's saved assets, map views, and stories, newest first, together with every portfolio on the account. Filter by kind, by portfolio, or by a free-text term over the label. Use this before manage_saved to get exact itemKeys and portfolio ids.
manage_alert
ChatGPTPause, resume, update, or archive one owned alert after the user explicitly requests that exact action. Resolve the alert id with list_alerts. For numeric updates, sanity-check changed thresholds against recent data. Archive is destructive; pause is reversible.
manage_saved
ChatGPTSave or unsave an asset, create a portfolio, or change what a portfolio holds. Only ever call this from an explicit user request — never to tidy up, and never on your own initiative after reading a list.
query_data
ChatGPTRun one bounded read-only ClickHouse SELECT against exact dotted tables returned by search_data or describe_data. The server rejects unregistered and out-of-scope tables, and every table must belong to the given iso.
query_grid_assets
ChatGPTSearch or resolve plants, zones, price nodes, substations, and transmission lines in one market. Provide exactly one of query or id. Copy a returned id unchanged into later tools; an LZ_ price target may correctly be a node: id.
read_project_file
ChatGPTA read document's passages with the locator a citation uses. Long documents come in pages: when truncated, call again with offset set to the passages already read.
request_asset_hydration
ChatGPTEnterprise: idempotently queue source-backed hydration for an existing grid asset. The request enters a reviewed publication workflow; it never publishes unreviewed research inline.
revoke_grant
ChatGPTRevoke one of your own OAuth grants by client id and resource. Call list_grants first to get the exact values.
screen_project_network
ChatGPTScreen many buses of a planning case at once, highest limit first, for choosing where to connect: the listed bus numbers, or every bus at minKv (default 100) and up, optionally filtered by name.
search_data
ChatGPTFind query-ready datasets in one market's catalog. Returns exact ClickHouse tables, grain, evidence class, and semantic query rules. Use describe_data for the selected dataset's full column schema before query_data.
search_project
ChatGPTRanked passages across every read document in a project.
search_wire
ChatGPTSearch collected US grid, market, and regulatory news, newest first. Covers ~225 publishers, ISO newsrooms, utilities and regulators. Optionally scope to one market (e.g. ercot, pjm) or sector (e.g. data-centers, battery-storage). Returns the publisher's canonical URL for every story — cite that, not this tool. Article bodies are never stored and never returned.
set_project_poi
ChatGPTPin the point of interconnection by bus number or name (as study results report it) and restudy every read planning case in the project there, on the server.
study_project_network
ChatGPTA DC screen of the MW a bus can inject (or withdraw) before a circuit reaches its rating, every outage studied. Studies the project's POI when bus is omitted. A screening estimate, not an interconnection study: report limits with the number. Nothing is saved.
