io.github.ecestnrkl/figjam-mcp

What it mainly works on, as our AI model read its tool descriptions: files and code on the machine it runs on; accounts and data in outside web services.

2 of 5 tools are described as changing something (creating, editing, deleting, sending or running).

• ingest_boardcreates or changes datadescribed as changing data

  “Read a FigJam/Figma URL with the configured Figma token and persist a local snapshot for get_board_context, answer_from_board and diff_board; the summary names the largest sections to read next and h…”

  Publisher declares: read-only no · destructive no · idempotent no · open-world yes

  7 inputs, 1 required

  • figmaFileUrlstringrequired“URL of the FigJam/Figma file to ingest”
  • figmaAccessTokenstring“Figma personal access token; falls back to FIGMA_ACCESS_TOKEN env var if omitted”
  • docStructureHintstring“Built-in framework used to map clusters to phases”
  • customPhasesarray“Free-form phase names to map clusters onto (e.g.”
  • ingestModestring“How aggressively to use vision LLM calls during ingest”
  • forceFullIngestboolean“Build the snapshot from the downloaded board again instead of the cached result for unchanged content, e.g.”
  • focusNodeIdsarray“IDs of pages, sections, frames, groups or single nodes whose image clusters are interpreted first, in list order; the rest of the board follows in the usual or…”
• get_board_contextreads or searches

  “Read source-cited context from an ingested board snapshot: original text, table rows, cluster summaries and arrows with at least one end, or a section, frame or group around one, on the page, each wi…”

  Publisher declares: read-only yes · destructive no · idempotent yes · open-world no

  8 inputs, 1 required

  • boardIdstringrequired“The Figma file key returned by ingest_board”
  • topicstring“Optional topic to focus the context on”
  • nodeIdsarray“Read exact source nodes, all rows of a table or the row of a table cell, or every excerpt inside a page, section, frame or group by its ID (nested sections inc…”
  • outsideSectionsboolean“Read every excerpt that lies in no section, cluster by cluster in reading order, tables row by row; on a board without sections, all excerpts.”
  • snapshotIdstring“Read a specific retained snapshot instead of the latest board”
  • limitinteger“Maximum evidence excerpts per page (default 20)”
  • cursorstring“nextCursor of the previous page, unchanged; send it with the same topic, nodeIds, outsideSections, limit and maxChars”
  • maxCharsinteger“Character budget for each representation: the readable text content and the JSON of structuredContent (default 12000, minimum 2000)”
• answer_from_boardreads or searches

  “Answer one specific question, such as a number, a status or a decision, from the most relevant excerpts of an ingested board, with validated source citations or an insufficient-evidence result.”

  Publisher declares: read-only yes · destructive no · idempotent no · open-world yes

  3 inputs, 2 required

  • boardIdstringrequired“The Figma file key returned by ingest_board”
  • questionstringrequired
  • snapshotIdstring
• diff_boardreads or searches

  “Compare retained local snapshots of the same board, reporting cluster, node and connector changes.”

  Publisher declares: read-only yes · destructive no · idempotent yes · open-world no

  2 inputs, 1 required

  • boardIdstringrequired“The Figma file key, as returned by ingest_board”
  • compareTointeger“How many snapshots back to use as the baseline (1 = the ingest before the latest)”
• diagnose_llm_configcalculates or convertsdescribed as changing data

  “Test the configured text and vision models with three small synthetic JSON challenges.”

  Publisher declares: read-only yes · destructive no · idempotent no · open-world yes

  Takes no inputs.

Each quoted sentence is the first sentence of the publisher’s own description. The kind of tool and the “changes data” mark are our AI model’s reading of that description, not something we tested: two differently worded runs had to agree, otherwise no label is shown, and nobody has checked them by hand. A tool without the “changes data” mark is described as only reading, searching or calculating, or the two runs did not agree. No tool was called. Read the publisher’s full description before relying on it. “Publisher declares” lists the hints the server itself sends with the tool (MCP annotations: read-only, destructive, idempotent, open-world); they are the publisher's claims, not checked by us. Inputs are the parameters in the tool's own input schema: name, declared type, whether it is required, and the first sentence of its description. Read 2026-10-23.

The registry verifies a package once, when it is published, and does not re-check it. A difference below is about the listing, not a judgement of the publisher.