Skip to main content
Glama
regulated-reporting-mcp

regulated-reporting-mcp

by dbett4

Regulated Reporting MCP

CI

The problem this answers

In regulated reporting work, three different things get collapsed into one “success”:

  1. the API accepted a request

  2. the change was actually applied

  3. somebody checked the result

Those are not the same event. A 202 is not “done.” An applied write is not “verified.” And a tool should not get to skip human confirmation just because a model asked nicely.

One-line pitch: give agents a small, guarded reporting API surface where writes need confirmation, results are read back, and receipts do not leak secrets or cell payloads.

This MCP server was built around that failure mode from government financial reporting work in Workiva-shaped systems. The default path stays safe: three guarded tools up front; the full tool catalog stays behind an explicit unsafe opt-in.

You can run the full path without Workiva credentials. The included mock covers the confirmation gate, rate-limit retry, asynchronous operation polling, readback, and receipt creation in under a minute.

Provenance. Sanitized public extract published August 2026; git history is publication history, not the original development timeline. Riverton demo and organizations are fictional — no client data, credentials, or private history appear here.

Mock-mode demo: list, read, gated write, verified readback, receipt

$ pip install -e . && workiva-mcp-demo

=== 5. Attempt the write WITHOUT confirm_write (contract blocks it) ===
{"error": "Tool requires confirmation: workiva_write_verified",
 "requires_confirmation": true, "risk": "write", ...}

=== 6. Re-issue with confirm_write=true (verified write + readback) ===
{"status": "verified", "written_cells": 3, "mismatches": [],
 "state": "verified", "receipt_uri": "workiva-receipt://2026...-workiva-write-verified-..."}

Related MCP server: production-grade-mcp-agentic-system

The problem it solves

Financial statements are unforgiving. A wrong number in a statement workbook is not a harmless UI bug. The default server therefore puts every mutation behind a reviewed tool contract and writes a redacted, create-once receipt. A tool that promises readback must report either verified or verification_failed. If it cannot tell, the dispatcher returns indeterminate.

What is implemented

  1. OAuth2 client credentials. The standard-library implementation caches a token, refreshes it before expiry, and clears and retries once after a 401 (auth.py, http_client.py).

  2. The less tidy parts of an enterprise API. The client handles 202 polling through both header and body operation locations, @nextLink and next_url pagination, parallel row-range reads, 429 backoff with Retry-After, and binary-safe xlsx/PDF exports (http_client.py, reapi/).

  3. A write gate that is enforced in code. All 117 registered tools have a reviewed effect / confirmation / proof contract. The compact dispatcher refuses tools with no contract, redacts mutation receipts, and treats missing readback as indeterminate (tool_contracts.py, compact_server.py, execution.py, receipt_store.py).

Architecture

flowchart LR
    C[MCP client] --> CS["compact server (3 tools):<br/>search / call / batch"]
    CS --> TC["tool contracts<br/>effect · confirmation · proof<br/>(fail-closed)"]
    TC --> FS["raw implementation registry<br/>(117-tool catalog; not served by default)"]
    FS --> T["tool families:<br/>cells · spreadsheets · documents ·<br/>files · linking · wdata · chains ·<br/>tasks · presentations · admin"]
    T --> H["http client<br/>OAuth2 · 401 retry · 429 backoff · pagination"]
    T -. selected mutation tools .-> EX["execution state machine<br/>202 ≠ applied ≠ verified"]
    EX --> H
    H --> W[(Workiva API)]
    H -. WORKIVA_MCP_MOCK=1 .-> M[(FakeWorkiva<br/>in-memory)]
    CS --> R["payload-redacted, create-once receipts +<br/>result artifacts<br/>(MCP resources)"]

The choices that matter:

  • For tools that use the mutation state machine, a 202 with no operation location is indeterminate, never success. execution.py accepts injected request and polling functions, so this behavior is testable without credentials. Not every raw tool uses that state machine; the test map says which claims apply where.

  • The manifest, not a guess based on the tool name, controls dispatch. scripts/generate_tool_contract_manifest.py --check verifies exact coverage of the 117-tool registry. Name-based classification is only a drift warning.

  • A receipt records an attempt and its reported outcome. It is not readback. proof: readback requires a deterministic verification result; otherwise the state is indeterminate and the receipt says pending_readback.

  • workiva-mcp and python -m workiva_mcp start the guarded three-tool server. Serving the raw catalog requires an explicit unsafe opt-in because direct calls bypass confirmation and receipt handling.

  • The three-tool front end reduces the listed tool count by about 97% and the listing payload by more than 90%, as measured by catalog_benchmark(). Large results become workiva-result://… MCP resources instead of chat payloads.

  • reapi/ is an offline-only layer for typed 4xx/5xx errors, retries, polling, and receipt validation. It performs no I/O and requires no authentication.

Quickstart (no credentials needed)

git clone https://github.com/dbett4/regulated-reporting-mcp
cd regulated-reporting-mcp
python -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"

workiva-mcp-demo        # end-to-end mock demo: list → read → gated write → verified readback
pytest                  # 126 credential-free tests
./scripts/proof.sh      # lint + manifest + tests + full offline demo

The demo uses FakeWorkiva, an in-memory transport with a synthetic "City of Riverton" ACFR-style workbook. It runs the same gate, one-shot 429 retry, 202 polling, readback, and receipt code used by the real transport.

The demo's verified write stores its receipt under ~/.workiva_mcp/receipts/ (the same write-once receipt store used against a real workspace). Set WORKIVA_MCP_RECEIPT_DIR to redirect receipts to any other directory.

Use it from Claude (mock mode)

{
  "mcpServers": {
    "workiva": {
      "command": "workiva-mcp",
      "env": { "WORKIVA_MCP_MOCK": "1" }
    }
  }
}

Against a real Workiva workspace

Provide OAuth2 client credentials (created in Workiva under an org API grant) and drop the mock flag:

export WORKIVA_CLIENT_ID=...
export WORKIVA_CLIENT_SECRET=...
export WORKIVA_REGION=us        # us | eu | apac
workiva-mcp                     # guarded 3-tool compact facade (default)

Credentials can also live in a .env file next to the package or pointed at with WORKIVA_ENV_FILE.

The raw 117-tool FastMCP server remains available for isolated development and for registry discovery inside the compact dispatcher. It bypasses confirmation and receipt handling, so it is deliberately awkward to start:

WORKIVA_MCP_CATALOG_MODE=full \
WORKIVA_MCP_ALLOW_UNGATED_FULL_CATALOG=1 \
workiva-mcp

Do not expose that mode to an autonomous or untrusted caller.

Tests

All 126 tests run without credentials. Transport tests use injected fakes; tests/test_mock_mode.py drives the actual tool stack through FakeWorkiva. Entrypoint tests check that compact mode is the default and that full-catalog mode fails unless both unsafe flags are present. Contract tests cover formula cells and exceptions after dispatch, and make sure a readback promise cannot be reported as applied_unverified.

See how to check each claim and the security notes.

Status

This is my public, sanitized implementation of patterns I used while building and tying out annual comprehensive financial reports in Workiva. It is not a copy of a client repository. It contains no engagement-management code, and all workbook values, identifiers, credentials, and organizations are synthetic.

License

MIT — see LICENSE.


Built by Dave Bettner.

Available Tools

117 tools
workiva_approve_taskB

Approve a task.

Args: task_id: The task UUID

Inputs: task_id (required). Returns: result

workiva_batch_edit_rich_textC

Batch edit rich text paragraphs.

Args: rich_text_id: The richText UUID revision: Current revision data: Edit operations

Inputs: rich_text_id (required), data, revision. Returns: result

workiva_bulk_edit_document_sectionsB

Bulk-edit sections within a document.

Args: document_id: The document UUID edits: Bulk edit operations

Inputs: document_id (required), edits. Returns: result

workiva_chain_runA

DEPRECATED — use workiva_run_chain.

Kept as a backward-compatibility shim; dispatches to the single-verb tool above.

Inputs: chain_id (required), environment_id, runtime_inputs. Returns: result

workiva_chainsA

DEPRECATED — use workiva_list_chains / workiva_get_chain / workiva_list_chain_runs / workiva_list_chain_run_nodes.

Kept as a backward-compatibility shim; dispatches to the single-verb tools above.

Inputs: filter, run_id, chain_id, operation, +1 more. Returns: result

workiva_copy_document_sectionA

Copy a document section within or to another document.

Args: document_id: Source document UUID section_id: Section UUID to copy target_document_id: Destination document UUID (empty = same document)

Inputs: section_id (required), document_id (required), target_document_id. Returns: result

workiva_copy_sheetA

Copy a sheet within the same spreadsheet or into a target spreadsheet.

Args: spreadsheet_id: The source spreadsheet UUID sheet_id: The sheet UUID to copy target_spreadsheet_id: Destination spreadsheet UUID (empty = copy within source)

Inputs: sheet_id (required), spreadsheet_id (required), target_spreadsheet_id. Returns: result

workiva_create_anchorC

Create a source anchor on spreadsheet cells.

Args: table_id: The table UUID (from workiva_spreadsheet_sheets) revision: Current table revision string start_row: Start row (0-based) stop_row: Stop row (0-based, same as start for single cell) start_column: Stop column (0-based, same as start for single cell) stop_column: Stop column (0-based, same as start for single cell)

Inputs: revision (required), table_id (required), stop_row, start_row, +2 more. Returns: result

workiva_create_document_sectionA

Create a new section within a document.

Args: document_id: The document UUID name: Section name (optional) index: Position index (-1 = append)

Inputs: document_id (required), name, index. Returns: result

workiva_create_fileB

Create a new file or folder in Workiva.

Args: name: File name kind: File type — Spreadsheet, Document, Presentation, or Folder container: Parent folder ID (optional)

Inputs: name (required), kind, container. Returns: result

workiva_create_sheetA

Create a new sheet in a spreadsheet.

Args: spreadsheet_id: The spreadsheet UUID name: Sheet name (optional — Workiva assigns one if omitted) index: Position index (-1 = append)

Inputs: spreadsheet_id (required), name, index. Returns: result

workiva_create_taskA

Create a new task.

Args: name: Task title (sent as title per the 2026-01-01 API; name is rejected) assignee: User ID to assign (wrapped as {id, type:"USER"}) approver: User ID for the approval step (wrapped as a participant) due_date: Due date ISO string file: File ID to associate description: Task description

Note: the v2026 Tasks API requires title (NOT name) and OBJECT assignees [{"id":..,"type":"USER"}] — confirmed against the live API 2026-05-31 with a readback-verified create probe. This tool maps the flat args to that shape so callers keep working.

Inputs: file, name, approver, assignee, +2 more. Returns: result

workiva_delete_document_sectionB

Delete a document section.

Args: document_id: The document UUID section_id: The section UUID

Inputs: section_id (required), document_id (required). Returns: result

workiva_delete_fileB

Return the explicit REST unsupported message for Workiva file deletion.

Inputs: file_id (required). Returns: result

workiva_delete_sheetA

Delete a sheet from a spreadsheet.

Args: spreadsheet_id: The spreadsheet UUID sheet_id: The sheet UUID to delete

Inputs: sheet_id (required), spreadsheet_id (required). Returns: result

workiva_delete_taskC

Delete a task.

Args: task_id: The task UUID

Inputs: task_id (required). Returns: result

workiva_document_sectionsA

DEPRECATED — use workiva_list_document_sections / workiva_get_document_section / workiva_create_document_section / workiva_delete_document_section / workiva_copy_document_section / workiva_bulk_edit_document_sections.

Kept as a backward-compatibility shim; dispatches to the single-verb tools above.

Inputs: document_id (required), name, edits, index, +3 more. Returns: result

workiva_duplicate_rich_textB

Duplicate a rich text resource.

Args: rich_text_id: The richText UUID

Inputs: rich_text_id (required). Returns: result

workiva_edit_table_structureA

Edit table structure — resize, autofit, hide/unhide, merge, insert/delete rows/columns.

Args: table_id: The table UUID operation: One of: resizeColumns, resizeRows, autofitColumns, autofitRows, hideColumns, hideRows, unhideColumns, unhideRows, mergeCells, unmergeCells, insertColumns, insertRows, deleteColumns, deleteRows params: Operation-specific parameters including "revision". Examples: resizeColumns: {"revision": "...", "columns": [0, 1], "width": 120} hideColumns: {"revision": "...", "columns": [2, 3]} mergeCells: {"ranges": [{"startRowIndex": 0, "endRowIndex": 1, "startColumnIndex": 0, "endColumnIndex": 4}]} insertRows: {"revision": "...", "index": 5, "count": 2}

mergeCells/unmergeCells take a `ranges` LIST of half-open index objects
{startRowIndex, endRowIndex, startColumnIndex, endColumnIndex} (end-EXCLUSIVE:
the example above merges A1:D1). A singular `range` is auto-wrapped into `ranges`.
These ops are ASYNC — the endpoint returns 202 with `operationLocation` in the
body; poll that URL until status == "completed" to confirm the merge applied.

Inputs: params (required), table_id (required), operation (required). Returns: result

workiva_export_documentB

Export a document.

Args: document_id: The document UUID format: Export format — pdf, docx, or xhtml options: Format-specific options (optional)

Inputs: document_id (required), format, options. Returns: result

workiva_export_fileA

Export a file from Workiva. Returns a download URL.

Args: file_id: The file UUID format: Export format (xlsx, pdf, csv, docx, xhtml)

Inputs: file_id (required), format. Returns: result

workiva_export_spreadsheetB

Export a spreadsheet.

Args: spreadsheet_id: The spreadsheet UUID format: Export format — xlsx, pdf, or csv

Inputs: spreadsheet_id (required), format. Returns: result

workiva_export_style_guideA

Export a style guide.

Args: style_guide_id: The style guide UUID

Inputs: style_guide_id (required). Returns: result

workiva_file_permissionsA

Get or modify file permissions.

Args: file_id: The file UUID operation: "get" to list permissions, "modify" to change them to_assign: Permissions to assign, e.g. [{"permission": "Edit", "principal": "user-id"}] to_revoke: Permissions to revoke, e.g. [{"permission": "Edit", "principal": "user-id"}]

Inputs: file_id (required), operation, to_assign, to_revoke. Returns: result

workiva_format_cellsA

Format cells via POST /content/tables/{tableId}/cells/edit (v2026).

table_id here is the sheet's OPAQUE content-table TOKEN (sheet.table.table — a long base64-ish string), NOT a 32-hex UUID. The legacy UUID path 404s on v2026. Colors are RGB objects {red, green, blue} with integer channels 0–255.

⚠️ v2026 PERSISTENCE (live-verified 2026-05-31 with before/after readback): NOT every op that returns 202 actually applies. On this cells/edit endpoint:

  • setCellsProperties (fillColor) — ✅ PERSISTS. Use this for cell fill.

  • setCellsBorders — borders (use NO_BORDER to clear; dedupe per edge).

  • formatCells (font/bold/color) — ⚠️ returns 202 but does NOT persist on v2026. For font/bold/size use the SEPARATE platform path: POST /spreadsheets/{ss}/sheets/{sid}/update with applyFormats.textFormat (fontFamily/fontSize/bold) — NOT exposed by this tool; call via api_request.

  • valueFormat / dataValidation / conditionalFormat — NO proven REST route on v2026 (UI-only); cells/edit returns 400 MalformedValue for DV/CF. A 202 from this tool is acceptance, NOT proof of apply — read the cell properties back to confirm.

Args: table_id: The opaque content-table token (from describe_workbook / get_sheet → table.table) operations: Array of format operations. Recommended types: - setCellsProperties (PERSISTS): {"type":"setCellsProperties","setCellsProperties":{"ranges":[...],"fillColor":{"red":230,"green":240,"blue":255}}} - setCellsBorders: {"type":"setCellsBorders","setCellsBorders":{"ranges":[...],"bottom":{"color":{"red":0,"green":0,"blue":0},"style":"single","weight":1}}} - formatCells (⚠️ 202 but no-persist on v2026 — see note): {"type":"formatCells","formatCells":{"ranges":[...],"color":{...},"fontFamily":"Arial","bold":true}}

Inputs: table_id (required), operations (required). Returns: result

workiva_get_chainC

Get chain details.

Args: chain_id: The chain UUID

Inputs: chain_id (required). Returns: result

workiva_get_documentA

Get document details.

Args: document_id: The document UUID expand: Comma-separated subresources (e.g., "sections")

Inputs: document_id (required), expand. Returns: result

workiva_get_document_sectionB

Get a single document section.

Args: document_id: The document UUID section_id: The section UUID

Inputs: section_id (required), document_id (required). Returns: result

workiva_get_fileA

Get details for a specific file by ID.

Args: file_id: The file UUID

Inputs: file_id (required). Returns: result

workiva_get_organizationC

Get organization details.

Args: org_id: Organization ID

Inputs: org_id (required). Returns: result

workiva_get_presentationA

Get presentation details.

Args: presentation_id: The presentation UUID

Inputs: presentation_id (required). Returns: result

workiva_get_presentation_slideC

Get a single presentation slide.

Args: presentation_id: The presentation UUID slide_id: The slide ID

Inputs: slide_id (required), presentation_id (required). Returns: result

workiva_get_sheetB

Get details for a single sheet.

Args: spreadsheet_id: The spreadsheet UUID sheet_id: The sheet UUID

Inputs: sheet_id (required), spreadsheet_id (required). Returns: result

workiva_get_spreadsheetB

Get spreadsheet details.

Args: spreadsheet_id: The spreadsheet UUID expand: Comma-separated subresources to include (e.g., "sheets,milestones")

Inputs: spreadsheet_id (required), expand. Returns: result

workiva_get_style_guideB

Get style guide details.

Args: style_guide_id: The style guide UUID

Inputs: style_guide_id (required). Returns: result

workiva_get_table_columnsB

Get column properties for a table.

Args: table_id: The table UUID

Inputs: table_id (required). Returns: result

workiva_get_table_propertiesC

Get table properties.

Args: table_id: The table UUID

Inputs: table_id (required). Returns: result

workiva_get_table_rowsB

Get row properties for a table.

Args: table_id: The table UUID

Inputs: table_id (required). Returns: result

workiva_get_taskA

Get details for a single task.

Args: task_id: The task UUID

Inputs: task_id (required). Returns: result

workiva_get_workspaceC

Get workspace details.

Args: workspace_id: Workspace ID

Inputs: workspace_id (required). Returns: result

workiva_import_fileB

Import a local file (xlsx/docx) into Workiva.

Args: name: Display name in Workiva file_name: Original filename (e.g., "report.xlsx") kind: Spreadsheet or Document file_path: Local path to the file to upload

Inputs: name (required), file_name (required), kind, file_path. Returns: result

workiva_import_style_guideC

Import a style guide.

Args: style_guide_id: The style guide UUID

Inputs: style_guide_id (required). Returns: result

workiva_list_chain_run_nodesA

List nodes within a chain run.

Args: chain_id: The chain UUID run_id: The chain run ID

Inputs: run_id (required), chain_id (required). Returns: result

workiva_list_chain_runsC

List runs of a chain.

Args: chain_id: The chain UUID

Inputs: chain_id (required). Returns: result

workiva_list_chainsC

List Wdata chains.

Args: max_results: Max results to return

Inputs: max_results. Returns: result

workiva_list_documentsA

List documents in the workspace.

Args: filter: OData filter (e.g., "name contains 'ACFR'") max_results: Maximum results to return

Inputs: filter, max_results. Returns: result

workiva_list_document_sectionsA

List sections within a document.

Args: document_id: The document UUID

Inputs: document_id (required). Returns: result

workiva_list_filesB

List files in the Workiva workspace.

Args: filter: OData filter (e.g., "name contains 'budget'", "kind eq 'Spreadsheet'") order_by: Sort order (e.g., "modified.dateTime desc", "name asc") max_results: Maximum files to return (default 50)

Inputs: filter, order_by, max_results. Returns: result

workiva_list_organization_rolesB

List roles in an organization.

Args: org_id: Organization ID

Inputs: org_id (required). Returns: result

workiva_list_organizationsB

List organizations.

Inputs: no named parameters. Returns: result

workiva_list_organization_solutionsA

List solutions enabled for an organization.

Args: org_id: Organization ID

Inputs: org_id (required). Returns: result

workiva_list_organization_usersB

List users in an organization.

Args: org_id: Organization ID

Inputs: org_id (required). Returns: result

workiva_list_presentation_layoutsC

List layouts in a presentation.

Args: presentation_id: The presentation UUID

Inputs: presentation_id (required). Returns: result

workiva_list_presentation_slidesA

List slides in a presentation.

Args: presentation_id: The presentation UUID

Inputs: presentation_id (required). Returns: result

workiva_list_sheetsA

List sheets within a spreadsheet.

Args: spreadsheet_id: The spreadsheet UUID

Inputs: spreadsheet_id (required). Returns: result

workiva_list_spreadsheetsA

List spreadsheets in the workspace.

Args: filter: OData filter (e.g., "name contains 'ACFR'") max_results: Maximum results to return

Inputs: filter, max_results. Returns: result

workiva_list_tasksA

List tasks in the workspace.

Args: filter: OData filter (e.g., "status eq 'Open'") max_results: Max results to return

Inputs: filter, max_results. Returns: result

workiva_list_workspace_groupsC

List groups in a workspace.

Args: workspace_id: Workspace ID

Inputs: workspace_id (required). Returns: result

workiva_list_workspace_membershipsA

List memberships in a workspace or workspace group.

Args: workspace_id: Workspace ID group_id: Group ID (optional — limits to a group's memberships)

Inputs: workspace_id (required), group_id. Returns: result

workiva_list_workspacesC

List workspaces.

Inputs: no named parameters. Returns: result

workiva_organizationA

DEPRECATED — use workiva_list_organizations / workiva_get_organization / workiva_list_organization_users / workiva_list_organization_roles / workiva_list_organization_solutions.

Kept as a backward-compatibility shim; dispatches to the single-verb tools above.

Inputs: org_id, user_id, operation. Returns: result

workiva_presentation_slidesB

DEPRECATED — use workiva_list_presentation_layouts / workiva_list_presentation_slides / workiva_get_presentation_slide / workiva_update_presentation_slide.

Kept as a backward-compatibility shim; dispatches to the single-verb tools above.

Inputs: presentation_id (required), patches, slide_id, operation. Returns: result

workiva_read_cellsA

Read cell values from a spreadsheet table. Returns data in TSV format.

Args: table_id: The table UUID (get from workiva_spreadsheet_sheets) start_row: First row to read (0-based, default 0) end_row: Last row to read (-1 = up to MAX_CELL_ROWS rows) start_col: First column to read (0-based, default 0) end_col: Last column to read (-1 = all columns)

Inputs: table_id (required), end_col, end_row, start_col, +1 more. Returns: result

workiva_read_rich_text_paragraphsC

Read paragraphs from a rich text resource.

Args: rich_text_id: The richText UUID

Inputs: rich_text_id (required). Returns: result

workiva_reapply_filtersC

Reapply filters on a file.

Args: file_id: The file UUID file_type: spreadsheet, document, presentation, or table

Inputs: file_id (required), file_type. Returns: result

workiva_reassign_taskC

Reassign a task.

Args: task_id: The task UUID

Inputs: task_id (required). Returns: result

workiva_reject_taskC

Reject a task.

Args: task_id: The task UUID

Inputs: task_id (required). Returns: result

workiva_rich_textB

DEPRECATED — use workiva_read_rich_text_paragraphs / workiva_batch_edit_rich_text / workiva_duplicate_rich_text.

Kept as a backward-compatibility shim; dispatches to the single-verb tools above.

Inputs: rich_text_id (required), data, revision, operation. Returns: result

workiva_run_chainB

Trigger a chain run.

Args: chain_id: The chain UUID environment_id: Environment ID (optional) runtime_inputs: Runtime input parameters as key-value pairs (optional)

Inputs: chain_id (required), environment_id, runtime_inputs. Returns: result

workiva_spreadsheet_sheetsA

DEPRECATED — use workiva_list_sheets / workiva_get_sheet / workiva_create_sheet / workiva_delete_sheet / workiva_copy_sheet.

Kept as a backward-compatibility shim; dispatches to the single-verb tools above.

Inputs: spreadsheet_id (required), name, index, sheet_id, +2 more. Returns: result

workiva_style_guideA

DEPRECATED — use workiva_get_style_guide / workiva_export_style_guide / workiva_import_style_guide.

Kept as a backward-compatibility shim; dispatches to the single-verb tools above.

Inputs: style_guide_id (required), operation. Returns: result

workiva_submit_taskC

Submit a task.

Args: task_id: The task UUID

Inputs: task_id (required). Returns: result

workiva_table_propertiesA

DEPRECATED — use workiva_get_table_properties / workiva_update_table_properties / workiva_get_table_columns / workiva_get_table_rows.

Kept as a backward-compatibility shim; dispatches to the single-verb tools above.

Inputs: table_id (required), patches, operation. Returns: result

workiva_task_actionsA

DEPRECATED — use workiva_submit_task / workiva_approve_task / workiva_reject_task / workiva_reassign_task.

Kept as a backward-compatibility shim; dispatches to the single-verb tools above.

Inputs: task_id (required), action. Returns: result

workiva_tasksA

DEPRECATED — use workiva_list_tasks / workiva_get_task / workiva_create_task / workiva_update_task / workiva_delete_task.

Kept as a backward-compatibility shim; dispatches to the single-verb tools above.

Inputs: file, name, filter, patches, +7 more. Returns: result

workiva_update_fileA

Update a file using JSON Patch operations (rename, move, etc.).

Args: file_id: The file UUID patches: JSON Patch array, e.g. [{"op": "replace", "path": "/name", "value": "New Name"}]

Inputs: file_id (required), patches (required). Returns: result

workiva_update_presentation_slideB

Update a presentation slide via JSON Patch.

Args: presentation_id: The presentation UUID slide_id: The slide ID patches: JSON Patch array

Inputs: slide_id (required), presentation_id (required), patches. Returns: result

workiva_update_table_propertiesC

Update table properties via JSON Patch.

Args: table_id: The table UUID patches: JSON Patch array

Inputs: table_id (required), patches. Returns: result

workiva_update_taskB

Update a task via JSON Patch.

Args: task_id: The task UUID patches: JSON Patch array

Inputs: task_id (required), patches. Returns: result

workiva_wdata_batch_refresh_connectionsA

Batch refresh multiple Wdata connections by fanning out to the per-connection endpoint.

The deprecated batch /connections/refresh endpoint (POST with {"connectionIds": [...]}) is dead — it returns 400 for every request shape. This function fans out to the working per-connection endpoint POST /connections/{id}/refresh (same as workiva_wdata_refresh_connection) and aggregates results.

Args: connection_ids: List of connection IDs to refresh

Inputs: connection_ids. Returns: result

workiva_wdata_cancel_query_runB

Cancel a Wdata query run.

Args: query_id: The query UUID run_id: The run ID

Inputs: run_id (required), query_id (required). Returns: result

workiva_wdata_connectionsA

DEPRECATED — use workiva_wdata_list_connections / workiva_wdata_get_connection / workiva_wdata_refresh_connection / workiva_wdata_batch_refresh_connections.

Kept as a backward-compatibility shim; dispatches to the single-verb tools above.

Inputs: operation, connection_id, connection_ids. Returns: result

workiva_wdata_create_queryB

Create a Wdata query.

Args: name: Query name sql: SQL statement

Inputs: sql, name. Returns: result

workiva_wdata_create_tableC

Create a Wdata table.

Args: name: Table name table_type: Table type schema: Table schema definition

Inputs: name, schema, table_type. Returns: result

workiva_wdata_delete_fileC

Delete a Wdata file.

Args: file_id: The Wdata file ID

Inputs: file_id (required). Returns: result

workiva_wdata_delete_queryC

Delete a Wdata query.

Args: query_id: The query ID

Inputs: query_id (required). Returns: result

workiva_wdata_delete_tableC

Delete a Wdata table.

Args: table_id: The Wdata table ID

Inputs: table_id (required). Returns: result

workiva_wdata_describe_queryB

Describe a Wdata query's output schema.

Args: query_id: The query ID

Inputs: query_id (required). Returns: result

workiva_wdata_filesA

DEPRECATED — use workiva_wdata_list_files / workiva_wdata_get_file / workiva_wdata_upload_file / workiva_wdata_delete_file.

Kept as a backward-compatibility shim; dispatches to the single-verb tools above.

Inputs: file_id, file_path, operation, sheet_name, +1 more. Returns: result

workiva_wdata_get_connectionC

Get a Wdata connection.

Args: connection_id: The connection ID

Inputs: connection_id (required). Returns: result

workiva_wdata_get_fileC

Get a Wdata file.

Args: file_id: The Wdata file ID

Inputs: file_id (required). Returns: result

workiva_wdata_get_queryB

Get a single Wdata query.

Args: query_id: The query ID

Inputs: query_id (required). Returns: result

workiva_wdata_get_query_resultB

Get the result of a Wdata query run.

Args: query_id: The query UUID run_id: The run ID

Inputs: run_id (required), query_id (required). Returns: result

workiva_wdata_get_tableC

Get a single Wdata table.

Args: table_id: The Wdata table ID

Inputs: table_id (required). Returns: result

workiva_wdata_list_connectionsB

List Wdata connections.

Inputs: no named parameters. Returns: result

workiva_wdata_list_filesB

List Wdata files.

Inputs: no named parameters. Returns: result

workiva_wdata_list_queriesC

List Wdata queries.

Args: max_results: Max results to return

Inputs: max_results. Returns: result

workiva_wdata_list_tablesA

List Wdata tables.

Args: max_results: Max results to return

Inputs: max_results. Returns: result

workiva_wdata_queriesA

DEPRECATED — use workiva_wdata_list_queries / workiva_wdata_get_query / workiva_wdata_create_query / workiva_wdata_delete_query / workiva_wdata_validate_query / workiva_wdata_describe_query.

Kept as a backward-compatibility shim; dispatches to the single-verb tools above.

Inputs: sql, name, patches, query_id, +2 more. Returns: result

workiva_wdata_query_runA

DEPRECATED — use workiva_wdata_run_query / workiva_wdata_get_query_result / workiva_wdata_cancel_query_run.

Kept as a backward-compatibility shim; dispatches to the single-verb tools above.

Inputs: query_id (required), run_id, operation. Returns: result

workiva_wdata_refresh_connectionC

Refresh a single Wdata connection.

Args: connection_id: The connection ID

Inputs: connection_id (required). Returns: result

workiva_wdata_run_queryC

Start a Wdata query run.

Args: query_id: The query UUID

Inputs: query_id (required). Returns: result

workiva_wdata_tablesA

DEPRECATED — use workiva_wdata_list_tables / workiva_wdata_get_table / workiva_wdata_create_table / workiva_wdata_delete_table / workiva_wdata_update_table.

Kept as a backward-compatibility shim; dispatches to the single-verb tools above.

Inputs: name, schema, patches, table_id, +3 more. Returns: result

workiva_wdata_update_tableB

Update a Wdata table via JSON Patch.

Args: table_id: The Wdata table ID patches: JSON Patch array

Inputs: table_id (required), patches. Returns: result

workiva_wdata_upload_fileA

Upload a local file to Wdata.

Args: file_path: Local file path (absolute)

Inputs: file_path (required). Returns: result

workiva_wdata_validate_queryA

Validate a Wdata query by ID or raw SQL.

Args: query_id: The query ID (optional — validates an existing query) sql: Raw SQL to validate (if query_id not provided)

Inputs: sql, query_id. Returns: result

workiva_workspaceA

DEPRECATED — use workiva_list_workspaces / workiva_get_workspace / workiva_list_workspace_groups / workiva_list_workspace_memberships.

Kept as a backward-compatibility shim; dispatches to the single-verb tools above.

Inputs: group_id, operation, workspace_id. Returns: result

workiva_write_cellsA

Write values to cells via the v2026 values API.

PUT /spreadsheets/{spreadsheet_id}/sheets/{sheet_id}/values/{A1Range}. Formulas pass through as text starting with "=". Cells are grouped by row and written as one PUT per row with empty strings filling any column gaps.

Args: spreadsheet_id: The spreadsheet UUID sheet_id: The sheet UUID (from workiva_spreadsheet_sheets) cells: [{row, col, value}] with 0-based row/col

Inputs: cells (required), sheet_id (required), spreadsheet_id (required). Returns: result

workiva_write_verifiedA

Write values/formulas, read back each row range, and report mismatches.

Formula cells are written, but Values API read-back may return calculated results rather than raw formula text. Those cells are reported separately as unverified so callers know when table-cell rawValue inspection is still needed.

Args: spreadsheet_id: The spreadsheet UUID sheet_id: The sheet UUID (from workiva_spreadsheet_sheets) cells: [{row, col, value}] with 0-based row/col retry_misses: Deprecated compatibility flag. Retries are intentionally disabled; mismatches are returned for an explicit, separately approved repair.

Inputs: cells (required), sheet_id (required), spreadsheet_id (required), retry_misses. Returns: result

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    C2.8/5.0

    Scored across 117 tools

    Disambiguation2/5

    Many deprecated umbrella tools (workiva_organization, workiva_tasks, workiva_wdata_files) sit alongside their single-verb replacements, creating overlapping entry points. File/document/spreadsheet tools also expose duplicate views of the same resources (export_file vs export_document vs export_spreadsheet; get_file vs get_document vs get_spreadsheet), which raises misselection risk.

    Naming Consistency3/5

    The dominant pattern is readable and mostly consistent (workiva_<domain>_<verb>_<noun>), but it is broken by a large set of deprecated resource-only shims like workiva_chains, workiva_rich_text, and workiva_spreadsheet_sheets. Exceptions such as workiva_full_link_flow and workiva_write_verified further reduce predictability.

    Tool Count1/5

    117 tools is far beyond what an agent can reasonably navigate, especially since roughly 19 are deprecated dispatch shims and many others duplicate the same underlying resources through different views. This creates severe context bloat and tool-selection overhead.

    Completeness4/5

    The server covers most core Workiva lifecycles: files, spreadsheets, documents/sections, Wdata tables/queries/connections, chains, tasks, links, and style guides. Minor gaps exist in presentation management (no list/create/delete), explicit file deletion is unsupported, and document creation/deletion is indirect, but agents can generally work around these.

    Maintenance

    ActivityStale
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      A
      quality
      D
      maintenance
      MCP server for deterministic numeric verification and spreadsheet auditing. Enables AI agents to verify math/finance claims, audit rows, and evaluate covenant rules with signed, independently checkable receipts.
      4
      MIT
    • A
      license
      Not graded
      quality
      D
      maintenance
      MCP server for regulated enterprises, providing per-tool RBAC, redacted audit logging, and structured error handling. Exposes bank tools for customer lookup, statement search, and dispute resolution over stdio and HTTPS transports.
      MIT