"Canvas" matching MCP tools:

• fluxink-mcp-chatgpt-show-handwriting-canvas

  fci-mcp

  Open the FluxInk handwriting recognition canvas. The user draws freehand strokes with a stylus, finger, or mouse. The strokes are converted by one of two model families: general recognition for handwriting, math, and chemical formulas, or structure recognition for molecular structures. Use this when the user asks to handwrite, draw, sketch, ink, scribble, or scrawl something. Use this when the user wants to draw a math equation, chemical formula, or molecular structure rather than type it. Use this when the user asks for a canvas, drawing pad, handwriting input box, or whiteboard. Use this when the user wants to convert stylus or finger drawings into recognized text or markup. Do NOT use this when the user types a question, equation, or formula in chat and just wants an answer. Do NOT use this when the user uploads or references an existing image of handwriting (call recognize_image instead). Do NOT use this when the user wants a formatted document, study sheet, or layout PDF (call create_layout instead). Do NOT use this when the user wants text rendered in a personal handwriting style (call show_style_canvas instead). Do NOT use this for conversational or informational requests that need no ink input. Do NOT re-open if a FluxInk handwriting canvas is already visible from any earlier turn. Instead instruct the user to keep drawing on the existing canvas. Only set force_new=true when the user explicitly asks for a brand new, fresh, or blank canvas. Always pass the original chat message in the prompt parameter so context is preserved after recognition. After calling, write a single short acknowledgement and do NOT describe the canvas UI.

  Connector

• fluxink-mcp-chatgpt-show-style-canvas

  fci-mcp

  Open the FluxInk natural texture ink style capture canvas. The user draws a few sample characters in a personal handwriting style. The widget then renders the requested text as a PNG spritesheet IN THAT EXACT PERSONAL HANDWRITING STYLE. Output is a downloadable spritesheet of individual handwritten glyphs that can be used as an asset (for example with image generation). Use this when the user asks to render or generate text in a personal handwriting style. Use this when the user wants a personalized handwritten note, card, letter, invitation, journal entry, or signature line that should look hand written by the user. Use this when the user wants to capture, clone, or sample a personal handwriting style and reuse it. Do NOT use this when the user just wants handwriting recognized (call show_handwriting_canvas instead). Do NOT use this for generic decorative handwritten fonts, calligraphy art, or AI generated script unrelated to the personal writing of the user. Do NOT use this to read text from an existing photo of handwriting (call recognize_image instead). Do NOT use this when the user wants a formatted document or layout (call create_layout instead). Do NOT use this for plain informational requests. Supports English and Chinese. Always pass the COMPLETE target text in the text parameter. Never truncate or abbreviate. Every character must appear in the spritesheet. Do NOT re-open if a FluxInk natural texture ink canvas is already visible from any earlier turn. Instead instruct the user to update the target text inside the existing widget. Only set force_new=true on an explicit user request for a brand new style canvas. After calling, write a single short acknowledgement and do NOT describe the UI. Once the user saves the spritesheet they can re-upload it for further design work.

  Connector

• eurostat_query_dataset

  eurostat-mcp-server

  Fetch statistical data from a Eurostat dataset with dimension filters. Returns decoded observations with dimension codes and labels, numeric values, an OBS_FLAG status (e.g., "p" = provisional, "e" = estimated) and a separate CONF_STATUS confidentiality marker (e.g., "C" = confidential, which is usually why a value is null), capped at 5,000 inline rows. Call eurostat_get_dataset_info first to discover valid dimension codes and values. Apply filters to keep the result set manageable — large unfiltered queries may trigger an async response error. Use filters.geo for specific country/region codes, or geo_level for NUTS hierarchy filtering (mutually exclusive). Use last_n_periods for the N most recent periods without knowing the end date. This tool fetches a slice: past the inline cap, either narrow the filters, or — on a deployment that runs a dataframe canvas — read the staged SQL table this response names in tableName with eurostat_dataframe_query rather than re-querying Eurostat. When the target is a whole dataset rather than a slice, eurostat_download_dataset reads the SDMX bulk endpoint instead and is the cheaper route.

  Connector

• describe_live_connector

  Power Automate MCP Server by Flow Studio

  Describe a live Power Platform connector/API and its operations. Aligned with Canvas MCP describe_api. Use mode=summary (default) to get a compact operation catalog. Use operationId to describe one operation, including inputs, outputs, dynamic parameter metadata, nextTool hints to call get_live_dynamic_options, and an authored hint + canonical example shape (when one exists) to copy into update_live_flow. Use search without connectorName/apiName to search operations across connectors and get connection-aware suggestions. Use mode=full only when raw OpenAPI connector metadata is required.

  Connector

• secedgar_get_fund_holdings

  secedgar-mcp-server

  List what an ETF or mutual fund holds, parsed from the NPORT-P portfolio report it files with the SEC every quarter. The input is the fund — a ticker like VOO, a fund series ID, or the registrant trust — which is the opposite direction from the ownership tools: secedgar_get_institutional_holdings and secedgar_find_holders answer who owns a company, this answers what a fund owns. Each position carries the security name, CUSIP/ISIN/LEI where the filer reports them, share balance, market value in USD, and percent of the fund's net assets, alongside fund-level net assets and total assets. Positions are returned largest-first by percent of net assets, one page of limit rows starting at offset; the full report registers as df_<id> when a canvas is available, which is how a fund running to thousands of positions is aggregated or joined against the 13F and insider dataframes. An NPORT-P covers exactly one fund series and a registrant trust files one report per series, so a trust with several funds needs the specific fund named — pass its ticker or series_id. Reports publish roughly two months after the period they cover, so every result is dated: the holdings are the portfolio as of report_period_date, not as of today.

  Connector

• openmeteo_get_forecast

  open-meteo-mcp-server

  Weather forecast for coordinates: hourly and/or daily variables for up to 16 days ahead, with optional past_days (up to 92) for recent history. Use past_days instead of openmeteo_get_historical for dates within the last 1–5 days, since ERA5 has a variable lag. Returns per-timestamp records — each hourly entry contains a "time" field (ISO 8601) plus one key per requested variable; each daily entry contains a "time" field (YYYY-MM-DD) plus requested variables. Common hourly variables: temperature_2m, precipitation, wind_speed_10m, relative_humidity_2m, cloud_cover, uv_index, apparent_temperature, precipitation_probability, weather_code, surface_pressure, visibility, wind_direction_10m, wind_gusts_10m, dew_point_2m. Common daily variables: temperature_2m_max, temperature_2m_min, precipitation_sum, wind_speed_10m_max, sunrise, sunset, uv_index_max, precipitation_hours, weather_code. A wide window — a large past_days plus many hourly variables — produces thousands of records; these spill to DataCanvas for SQL querying when canvas is enabled, and return a bounded preview with truncated: true when it is not. At least one of hourly_variables or daily_variables is required.

  Connector

• openmeteo_get_flood

  open-meteo-mcp-server

  GloFAS (Global Flood Awareness System) river discharge forecast and historical reanalysis. Returns daily ensemble river discharge (m³/s) for the river nearest to the given coordinates — no river ID needed, the API snaps to the nearest stream. Forecast horizon up to 210 days ahead; reanalysis history back to 1984-01-01. One mode per call: forecast_days for the future outlook, or start_date and end_date together for reanalysis history. The two modes are mutually exclusive, and a date range needs both ends — a lone start_date or end_date is rejected. Available daily variables: "river_discharge" (ensemble mean), "river_discharge_mean", "river_discharge_min", "river_discharge_max", "river_discharge_median", "river_discharge_p25" (25th percentile), "river_discharge_p75" (75th percentile). Returns null for coordinates far from any river or in areas without GloFAS coverage. A wide reanalysis range produces thousands of daily records and spills to DataCanvas for SQL querying when canvas is enabled, returning a bounded preview with truncated: true when it is not.

  Connector

• openmeteo_get_climate

  open-meteo-mcp-server

  Long-range climate projections from bias-corrected daily CMIP6 models, covering 1950-01-01 to 2050-12-31 at any coordinate. Answers "what will conditions look like through 2050?" — the future-projection counterpart to openmeteo_get_historical (ERA5, what happened). Daily resolution only. Available models: CMCC_CM2_VHR4, FGOALS_f3_H, HiRAM_SIT_HR, MRI_AGCM3_2_S, EC_Earth3P_HR, MPI_ESM1_2_XR, NICAM16_8S. A model name outside that list is sent upstream rather than rejected here, so a model Open-Meteo adds later still works; if upstream rejects the request, the error names the offending model on its own rather than the whole requested list. With 2+ models each variable appears once per model with the model name as suffix (e.g. temperature_2m_max_CMCC_CM2_VHR4); a single or omitted model returns plain variable names. Not all models carry all variables — missing combinations return null. Multi-decade daily pulls across several models produce thousands of records and spill to DataCanvas for SQL querying when canvas is enabled, returning a bounded preview with truncated: true when it is not.

  Connector

• met_search_collections

  met-museum-mcp-server

  Search the Metropolitan Museum of Art collection by keyword and optional filters. Returns the total match count and a page of matching object IDs, which met_get_object resolves to full records. Relevance is keyword-based, not semantic; department and geographic filters narrow results more than a longer query. The medium parameter maps to the classification field (pass "Paintings", "Drawings", etc., not material descriptions like "Oil on canvas"). isPublicDomain guarantees CC0-licensed images; hasImages also includes copyrighted works. isOnView restricts results to works currently on display in a Met gallery.

  Connector

• eurostat_dataframe_query

  eurostat-mcp-server

  Run a read-only SQL SELECT against tables staged on a Eurostat dataframe canvas — the way to reach observations past the 5,000-row inline cap of eurostat_query_dataset and past the inline preview of a eurostat_download_dataset bulk download, and to aggregate, group, or join across staged tables without re-fetching from Eurostat. Call eurostat_dataframe_describe first for the table and column names, which differ between the two stagers. Only a single SELECT statement runs: statement chaining, non-SELECT verbs, and functions that read files or external data are rejected. Columns are flat — every dimension is a code column named after the dimension, the measure is obs_value, the observation flag is obs_flag / obs_flag_label and the confidentiality marker is conf_status / conf_status_label; a "_label" companion per dimension exists only on tables eurostat_query_dataset staged. Both stagers write the same five measure columns with the same codes, so join their tables on dimension codes and time and compare obs_flag or conf_status across them directly.

  Connector

• get_clip

  Clueso

  Read one clip: its elements (positions/sizes in canvas pixels), voiceover (text, voice, duration, voiceover_volume), background and transition. Pass `render` to also get a PNG of the frame. ASK FOR WHAT YOU NEED. A full read is large — on a dense clip the per-word voiceover array and the element type_data blobs dominate it, and repeated full reads are the main way a long session runs out of context. `select` returns exactly the parts you name: select: ['elements.x','elements.y','elements.width','elements.height'] → geometry only, to fix a layout select: ['elements.name','elements.start_time','elements.end_time'] → a timing pass select: ['words'] → word timings only, to sync visuals to narration select: ['elements.textdata','words'] → rewrite copy against the VO select: ['elements'] → whole element rows, no words select: [] → no JSON at all (pair with render for the PNG alone — smallest read) (omit select) → everything; fine for a first look, expensive to repeat `render` is the other output, and it is separate from `select`: `select` shapes the JSON, `render` produces a PNG. render: {} → the frame at t=0 render: { timestamp: 2.5 } → the frame 2.5s into the clip render: { save: true } → also uploads the PNG and returns presigned_url select: [], render: {} → the PNG alone, no JSON select: ['elements'], render: {} → element rows AND the frame Omitting `render` renders nothing. `timestamp` and `save` live inside it because they only mean anything for a render — there is no way to ask for them without asking for the image. `element_ids` is the other axis: it picks WHICH element rows come back, independently of `select`. Combine them for the leanest read — e.g. element_ids: ['el_9'], select: ['elements.x','elements.y']. Element shape: universal wrapper fields (id, geo, name, x, y, width, height, start_time, end_time, rotation) plus type-specific data (textdata/shapedata/imagedata/videodata/zoomdata) plus an optional `keyframes` array when animated. Keyframes come back in the same flat wire shape add_elements takes — { timestamp, positionX?, positionY?, width?, height?, interpolation? } in canvas pixels — so you can round-trip read → edit → update_elements without reshaping. Clip-level fields include `transition` (the current transition object — sibling of the update_clips `transition` arg; null if none) and `voiceover_words` (per-word timestamps; null on clips with no transcription).

  Connector

• update_clips

  Clueso

  Update one or more clips in a single call — background, transition, duration, title, visibility, position, and (video clips) the footage transform via video_styling. Always batch. Each entry: { clip_index, duration?, title?, background_color? | background?, transition?, hidden?, position?, voiceover_volume?, video_styling? } - voiceover_volume: 0-100. This is how you make a clip SILENT — set 0 to mute the narration while keeping the text and its word timings. get_clip and get_project report the current value, so read it before muting to restore it later. There is no "delete voiceover" tool; muting is the supported way. - background_color: hex (keeps the existing pattern/image, changes color). background: full object from find(type='backgrounds') — replaces the whole background. - transition: { type, duration, direction?, color? }, or null to remove. Set on the OUTGOING clip — the transition plays at the END of clip_index, into the next clip. Types: circleWipe, rainbowWipe, cornerWipe, stack, colorWipe, dissolve, pushPull, fadeToColor. Directions: rainbowWipe/cornerWipe take topLeft|topRight|bottomLeft|bottomRight; colorWipe left|right; stack/pushPull left|right|top|bottom. - position: move the clip to a new zero-based index. Moves apply FIRST (in array order), then property patches — keep reorders in a separate call from property edits so indices stay predictable. - video_styling (VIDEO CLIPS ONLY — ignored on slide clips): reposition/resize the clip's recording on the canvas, e.g. into a designed window slot. { position?: [cx, cy], size?: [w, h] } as canvas fractions 0–1 ([1,1] = full-bleed, position is the footage CENTER), plus crop? and border?. Fields merge onto the clip's existing video_styling. Concurrency: parallel-safe per clip — all property updates land in ONE save. Safe to fan out across DIFFERENT clips. Do NOT run concurrently with element/voiceover edits on the SAME clip, or whole-project mutations (add_clips, add_audio, update_project). When adding audio + resizing clips, run update_clips BEFORE add_audio.

  Connector

• generate_video

  Hermoso

  Render a RAW video clip from your own prompt and return its served mp4 URL. For finished brand ADS prefer render_ad (it runs the Studio quality pipeline — composited text, clean speech, end card, music); use this for raw/experimental clips or precise manual control. ONE generation = one continuous clip up to the model’s longest listed duration — the longest-clip model in the catalog today renders a full multi-beat spot of up to 30 SECONDS in ONE unbroken take with native synchronized audio, so never assume a generic 8–10s cap and never stitch something that fits one clip; durationSeconds must be one of the model’s durations from hermoso_capabilities, which is the live list. TO GET A SPECIFIC MODEL, NAME IT in `model`: an unnamed render is routed by the server’s own auto-pool, which is narrower than the catalog, so the longest-clip and highest-resolution models are reached by naming them and not by omitting the field. Renders take 1–3 min. refImage anchors the opening frame; ttsScript adds a voiceover. AUDIO IS NOT FREE AND NOT OPTIONAL BY DEFAULT: a clip delivered with no audio of its own gets a music bed composed and CHARGED on top of the render (see musicMood and audio) — on a cheap short draft the bed can cost as much as the clip. Pass refVideo (a clip URL) to EDIT an existing video instead of generating from scratch — the omni engine transforms that clip per your prompt, inheriting the source clip’s canvas + length (aspectRatio/durationSeconds are ignored for an edit). Spends credits (Starter plan is video-blocked server-side).

  Connector

• secedgar_get_insider_transactions

  secedgar-mcp-server

  Fetch Form 4 insider transactions (purchases, sales, grants, exercises) for a company by parsing SEC EDGAR ownership XML. Returns the reporting person, their relationship to the issuer, transaction date, type, shares traded (absolute magnitude), direction (acquire/dispose), price per share, and shares owned after the transaction. Covers nonDerivative transactions (open-market buys/sells, gifts) and derivative transactions (option exercises, RSU vests). When a canvas is available, the full set of transactions parsed from the scanned recent filings is materialized as df_<id> (the inline list is a preview capped at limit) — query it with secedgar_dataframe_query to aggregate net buy/sell by insider: SUM(CASE WHEN direction='dispose' THEN -shares_traded ELSE shares_traded END). Use secedgar_search_filings with forms=["4"] for broader date-range queries or to search across all companies.

  Connector

• secedgar_get_institutional_holdings

  secedgar-mcp-server

  Fetch 13F-HR quarterly institutional holdings by parsing the SEC EDGAR information table XML. ticker_or_cik is the institutional filer — its 10-digit CIK (e.g. 0000102909), or an entity name resolved through EDGAR entity search — and the tool returns what that institution holds. A name that matches several EDGAR filers (some legal names are shared across entities) returns those candidates so you can retry with the exact CIK, rather than guessing. For the reverse direction — which institutions hold a given portfolio company — use secedgar_find_holders, whose filer_cik results feed straight back into this tool. The 13F information table lists each position: issuer name, CUSIP, shares held, market value (in whole USD), and put/call designation for options. Sub-lines for the same security are consolidated into distinct positions sorted by value by default (set consolidate=false for raw filing rows). The inline holdings list is one page of limit rows starting at offset — pass the returned next_offset to walk further down a large information table. The full parsed holdings set is also materialized as df_<id> when a canvas is available — so query it with secedgar_dataframe_query to aggregate the whole filing or self-join across quarters on cusip + reporting_period. Institutions with less than $100M in 13(f) securities are exempt and may not file. Use secedgar_search_filings with forms=["13F-HR"] for broader search.

  Connector

• bls_dataframe_query

  bls-labor-mcp-server

  Run a single-statement SELECT against the canvas dataframes registered by bls_get_series. Read-only: writes, DDL, DROP, COPY, PRAGMA, ATTACH, and external-file table functions are rejected. System catalogs (information_schema, pg_catalog, sqlite_master, duckdb_*) are denied at the bridge layer — use bls_dataframe_describe to list available dataframes. Supports JOINs, aggregates, window functions, and CTEs. Optional register_as persists the result as a new dataframe with a fresh TTL for chained analysis. Canvas SQL operations consume zero BLS API quota. Requires CANVAS_PROVIDER_TYPE=duckdb.

  Connector

• eia_dataframe_describe

  eia-energy-mcp-server

  List canvas dataframes (df_<id>) materialized by eia_query_route calls that passed stage: true, with provenance, expiry, row count, and column schema. Nothing is staged until such a call runs, so an empty list on a fresh session means no query has staged yet, not that staging failed. Drops entries for dataframes the canvas no longer holds before responding, so the list is always current. Pass a specific name to inspect one dataframe; omit to list all active dataframes for this tenant. A name that is not staged comes back as found=false alongside the handles that are, never as an empty list. Listing is not use: only an eia_dataframe_query statement naming a dataframe extends its expiry, so a dataframe polled with this tool and never queried still lapses on schedule.

  Connector

• water_find_sites

  usgs-water-mcp-server

  Find USGS water monitoring sites by bounding box, state, county, or HUC watershed code, filtered by site type and parameter availability. Returns site numbers, names, coordinates, types, altitude, and (in expanded mode) drainage area. Call this first — water_get_readings, water_get_series, and water_get_conditions all require a site number. Capped at 500 sites inline; when truncated=true, upstreamTotal holds the full count and, if DataCanvas is enabled, the complete match set stages to a canvas (canvas_id/table_name) for retrieval via water_dataframe_query — otherwise narrow the filters to get all matches.

  Connector

• eurostat_download_dataset

  eurostat-mcp-server

  Download a Eurostat dataset in bulk through the SDMX 2.1 TSV endpoint and stage every observation as a SQL table on the dataframe canvas — the route to a whole dataset, where eurostat_query_dataset is the route to a slice of one. The TSV wire format is roughly half the bytes of the JSON-stat body eurostat_query_dataset reads, so it reaches datasets that would otherwise time out, and it is expanded here into one row per observation. Filters take the same dimension-code map eurostat_query_dataset uses and are applied server-side by Eurostat; call eurostat_get_dataset_info first for the dimension codes and eurostat_get_dimension_values for their values. Narrow with since_period/until_period rather than asking for the most recent N periods — the TSV layout keeps a column for every period whichever is requested, so a period range is what actually shrinks the response. Transfers are bounded by a byte budget enforced while streaming: when it is spent the download stops and budgetExceeded is set, leaving a prefix of the dataset rather than an error. Only preview_limit rows come back inline; the rest are reachable with eurostat_dataframe_query when this deployment runs a canvas, and are not retained when it does not.

  Connector

• get_self_view

  TunnelMind Data API

  P83 Gate 1. The caller is the subject: every fact here comes from the TLS handshake and headers the caller has already sent, so there is no render, no browser, and nothing to authorize. Three surfaces: - `client` — user-agent, HTTP protocol, TLS version/cipher, ClientHello length, the pre-hashed JA3/JA4 input components, header order. - `state` — how many cookies were sent (never their values), Referer, DNT, Global Privacy Control. - `network` — address, ASN and operator, coarse geography, edge colo, and the four-lens verdict on the caller's own IP. Every field carries a coverage state from the same three-value vocabulary as `/v1/verify`: `observed_clean`, `never_observed`, `degraded`. There is no fourth state. Fields that run inside a page — localStorage, canvas fingerprinting, cookie values — are reported `never_observed` with reason `not_observable_server_side`, because they are outside a server's vantage rather than missing. `claim_vs_conduct` compares the claimed user-agent against the shape of the request itself and returns `consistent`, `mismatch`, or `unverifiable`, with the evidence listed. It is deliberately narrow: JA4 requires Cloudflare Enterprise + Bot Management, so there is no portable fingerprint to look up in a public corpus, and this check only reports contradictions it can demonstrate from the request in hand. `unverifiable` is the honest default and is never dressed up as a pass. Use this tool when: - You want to know what a server learns about your client without asking you anything. - You are checking whether a client's user-agent claim matches its conduct. Do NOT use this tool when: - You need facts about some OTHER host — that is `POST /v1/verify/{node}`. Must be called directly at `data.tunnelmind.ai`. Behind a proxy, the connection properties describe the proxy, not the caller. `?receipt=true` attaches a signed Receipt v1.0 committed to the transparency log.

  Connector