MCP App Store
by KeenEthics
The Cards icon

The Cards MCP Server for ChatGPT

By Icons Of

Overview

Sustainability reporting shouldn't feel like paperwork. The Cards brings your organization's ESRS and CSRD reporting into ChatGPT, so you can work with your sustainability data the way you'd talk to a colleague who knows it inside out. Ask where your report stands and get a real answer. Find out which topics are material for your business, what you've already disclosed on climate, workforce or governance, and what the standard actually expects from you – in plain language, not legal text. Need to know how your emissions disclosure compares to last year's? Just ask. When it's time to move, The Cards moves with you. Record answers, mark disclosures as material, hand off datapoints to the right colleague and close open threads – without leaving the conversation. Upgrading to a newer version of ESRS? Walk through the mapping together instead of starting from scratch. The Cards works as you: it sees your organization only, does nothing your own account couldn't, and keeps a log of every change. Your data stays yours. Sign in with your The Cards account to get started.

How to connect

  1. ChatGPT: Open the The Cards listing in the ChatGPT app directory and choose Connect. ChatGPT listing The Cards exposes 61 tools here.

Tools

cards_esrs_get_datapoint

ChatGPT
Returns the standard's definition of a single datapoint, including its guidance text, reporting type, unit and field/column definitions. reportingType is the CMS type; valueKind is the normalized kind (text, number, percent, boolean, date, enum, object, table). unitSource is 'cms' when the unit comes from structured CMS data, 'inferred' when it was guessed from the description, and null when there is no unit.

cards_esrs_get_topic

ChatGPT
Returns one topic from an ESRS version, including its explanatory text.

cards_esrs_list_datapoints

ChatGPT
Lists the datapoints the standard defines under one disclosure requirement.

cards_esrs_list_disclosures

ChatGPT
Lists the disclosure requirements the standard defines under one topic.

cards_esrs_list_topics

ChatGPT
Lists the topics defined by an ESRS version, such as E1 Climate change. Use this for questions about what the standard requires, rather than what a report contains. description is plain text; descriptionHtml is only present when the CMS source contained markup. eurLexUrl is the legal-text deep link when the CMS has base_url/doc_id.

cards_esrs_list_versions

ChatGPT
Lists the published ESRS standard versions. This is global reference data describing the standard itself, not organization content.

cards_gap_get_datapoint

ChatGPT
Returns the gap analysis recorded for a single datapoint in a report, including the full comment and CMS description. Reports a not_found error when no gap analysis has been filled in for that datapoint yet. dataCoverage is a percentage (0-100). remainingEffort is one of 'S', 'M', 'L', 'XL' or 'No work'. automationLevel is one of 'S', 'M', 'L' or 'XL'.

cards_gap_list_for_disclosure

ChatGPT
Returns the compact gap-analysis rows for one disclosure requirement, paginated. Same fields as cards_gap_list_for_report. Prefer that tool with disclosureId unless you already know the disclosure. dataCoverage is a percentage (0-100). remainingEffort is one of 'S', 'M', 'L', 'XL' or 'No work'. automationLevel is one of 'S', 'M', 'L' or 'XL'.

cards_gap_list_for_report

ChatGPT
Returns a compact gap-analysis overview for a report: summaryByTopic (counts, average coverage, effort and automation breakdowns, worst coverage first) plus a paginated page of gap rows. Only datapoints with a filled-in GAP record appear — this is planning data, not completion. CMS descriptions are omitted so the payload stays small; call cards_gap_get_datapoint for the full comment and requirement text. Filter with topicId or disclosureId, then paginate with limit/offset while hasMore is true. dataCoverage is a percentage (0-100). remainingEffort is one of 'S', 'M', 'L', 'XL' or 'No work'. automationLevel is one of 'S', 'M', 'L' or 'XL'.

cards_gap_set_datapoint

ChatGPT
Creates or updates the gap analysis for one datapoint in an active report. Pass only the fields that change; omitted fields keep their current value (or null on first write). dataCoverage is a percentage (0-100). remainingEffort is one of 'S', 'M', 'L', 'XL' or 'No work'. automationLevel is one of 'S', 'M', 'L' or 'XL'. This is planning data, not the recorded answer. Requires mcp:write:gap and write access. There is no per-field history table — the write is recorded in mcp_audit_log.

cards_identity_get_current_organization

ChatGPT
Returns the organization this MCP token is scoped to. Every other tool reads data from this organization only.

cards_identity_get_current_permissions

ChatGPT
Returns the caller's permission level per resource type (report, workflow, files, and so on), this token's MCP scopes, and any report restriction on the token. Useful for explaining why a read or write was refused.

cards_identity_get_current_user

ChatGPT
Returns the user this MCP token acts as. Use it to confirm whose data you are reading before answering questions about 'my' reports.

cards_organization_files_get_content

ChatGPT
Returns the extracted text of a library file or a datapoint attachment as markdown. Use this when you need the whole document rather than the matching passages from cards_search_organization_files. Large files are paginated with offset/limit; follow nextOffset while hasMore is true. Extractable types: PDF, Excel, Word and PowerPoint (.pdf, .xlsx, .xls, .docx, .pptx). Images and other binaries return metadata with extractable: false. Unindexed files can still be read here.

cards_organization_files_list

ChatGPT
Lists the reference files uploaded to the organization's internal library, with their descriptions, tags, reporting periods, whether they are indexed for semantic search, and whether their text can be extracted. Returns metadata only; call cards_organization_files_get_content for the extracted text.

cards_organization_files_list_datapoint_attachments

ChatGPT
Lists files attached to datapoints in reports the caller can read, showing which report and datapoint each one supports, whether it is indexed for semantic search, and whether its text can be extracted. Optionally narrow to a single report or datapoint. Call cards_organization_files_get_content to read a file.

cards_organization_get_profile

ChatGPT
Returns the collected company profile from onboarding — the same markdown/chat context the app's internal AI uses when suggesting datapoint answers, plus the TIC.io company briefing. Answers are prose (answer, reasoning, chatHistory), not a structured snapshot. An organization may have several profiles (for example a parent and subsidiaries); omit profileId to get the most complete current-onboarding profile. Retired structured company-profile experiment runs are not included.

cards_publication_get

ChatGPT
Returns one publication with its full settings: status, publishing window, language, theme, introduction text and public link options.

cards_publication_list

ChatGPT
Lists the organization's publications — the shareable, themed public views of a report — with their draft/published status and which report each one publishes. Use this to answer what has been published and whether a given report is publicly visible.

cards_report_accept_copy_item

ChatGPT
Writes a reviewed mapping into the copied report and closes that queue item. Omit value/fields/rows to accept the proposed answer as-is (or an empty scalar if there is no proposal — empty is valid). To use your own wording instead, fill in the item's write example (write.argument, write.example). Text is plain text or markdown; HTML is refused. contentOrigin is ai; actor.type is mcp_token. Requires mcp:write:reports and write access to an active report.

cards_report_can_write

ChatGPT
Checks whether this token may write to a report. It does not change any data. Returns an error if the token lacks a write scope, the user cannot write the report, or the report is not active (archived or finalized). Use this before a write tool, or to confirm why a write was refused.

cards_report_copy

ChatGPT
Creates a new report from an existing one: either a roll-forward on the same standard or a migration onto a mapped newer version. Matching answers are copied; anything uncertain is left in a review queue (cards_report_list_copy_items). Pick targetEsrsVersionId from cards_report_list_copy_targets; omit it to stay on the source standard. Requires mcp:write:create_reports, permission to create reports, and write access to the source (which may be finalized). A token limited to selected reports cannot create new ones. This is not cards_report_create, which starts empty.

cards_report_create

ChatGPT
Creates a new sustainability report in the token owner's organization. Pick esrsVersionId from cards_esrs_list_versions (the row with activeVersion true is the current standard). The report starts active and empty. Requires mcp:write:create_reports; the user must already be allowed to create reports in the app. A token limited to selected reports cannot create new ones. To copy answers from an existing report, use cards_report_copy.

cards_report_create_comment

ChatGPT
Adds a comment on a datapoint in a report. Omit threadId to start a new thread; pass threadId to reply. The comment is stored as the token's user (it looks like them in the app). To notify someone, pass mentionUserIds — they are stored as the same @[Name](userId) markup the app uses. You may also include that markup in text. Does not delete anything. Requires mcp:write:comments and write access to an active report.

cards_report_get

ChatGPT
Returns the metadata for one report: name, status, reporting period, ESRS version and responsible user. previousReport is the readable report in the same organization with the latest earlier periodTo (same ESRS version preferred). There is no stored previousReportId — this is inferred from periods.

cards_report_get_copy

ChatGPT
Returns the version-mapping job behind a report that was copied from an earlier report or onto a newer ESRS version: source report, source and target standards, how many previous answers still need a decision, and a review URL. Use this first, then cards_report_list_copy_items to walk the queue. Reports that were not created by a copy return not_a_copy. Requires mcp:read:reports.

cards_report_get_datapoint

ChatGPT
Returns the recorded answer for a single datapoint in a report, together with its reporting type, unit, field/column definitions and structured rows when the datapoint is a table. reporting is true when an input field is shown; completed means it is marked done. Datapoints turned off in the app (active is false) are not returned — the tool reports datapoint_inactive. reporting false means this row is not fillable — ESRS scope totals are often written on the parent (E1-6-44) while children such as E1-6-44-a are label-only. Use valueKind to tell numbers, booleans, enums and tables apart; reportingType is the CMS type (Text, Number, Percent, Boolean, Date, Json, JsonArray, JsonTable). unitSource is 'cms' when the unit comes from structured CMS data, 'inferred' when it was guessed from the description, and null when there is no unit. description is plain text; descriptionHtml is only present when the CMS source contained markup. Pass includeComparative=true to also return the same datapoint from the previous reporting period (inferred from periodTo). For previous values of this datapoint in the same report, use cards_report_get_datapoint_history. To read the same datapoint across every report you can access, use cards_report_get_datapoint_across_reports.

cards_report_get_datapoint_across_reports

ChatGPT
Returns the recorded answer for one or more datapoints in every report the caller can read. Use this for questions such as 'which reports have a scope 1 value?' instead of calling cards_report_get_datapoint once per report. Pass every identifier that might apply, for example both ESRS E1-6-44-a/b/c and VSME B3-30-a/b/c. Each row includes reporting (true when an input field is shown), completed (marked done) and hasContent (a value is stored). Datapoints turned off in a report (active is false) are omitted. ESRS scope totals are often on the parent E1-6-44 (reporting true) while children such as E1-6-44-a have reporting false; do not treat those children as unanswered. Reports whose ESRS version does not include the datapoint are omitted. hasContent is true only when an actual value is stored.

cards_report_get_datapoint_history

ChatGPT
Returns previous versions of a datapoint's stored answer in one report, newest first. Each history row is a snapshot taken when the value was next overwritten (the live value is current, not a history row). versionId uniquely identifies a revision for later restore. actor.type is user, mcp_token or app_ai; lastUpdatedBy is the human whose permission was used. contentOrigin is ai/user/mixed when known. restoredFromVersionId is set when that version was itself a restore. fromCopy is true when that snapshot was carried over from a copied report. If this report was created by copying another, copiedFrom names the source field and the value it had at copy time — that is how the mapping happened. Same report read access as cards_report_get_datapoint. Inactive datapoints are not returned (datapoint_inactive). Paginate with limit/offset; follow counts when hasMore is true. This tool does not change any values.

cards_report_get_datapoint_suggestions

ChatGPT
Returns the latest stored AI suggestion for one datapoint in a report, including timestamp, model, reasoning and sources. It never triggers generation. To read the value the user actually saved, use cards_report_get_datapoint (see contentOrigin there).

cards_report_get_disclosure

ChatGPT
Returns one disclosure in a report together with its datapoints: completion state, reporting type, unit and field/column definitions. Recorded answers are not included; call cards_report_get_datapoint for a specific answer. Each datapoint includes reporting (true when an input field is shown), completed (marked done) and hasContent (a value is stored). Datapoints turned off in the app (active is false) are omitted — they are not part of the report. reporting false means the row is label-only — ESRS scope 1/2/3 totals are often written on the parent (E1-6-44, reporting true) while children such as E1-6-44-a have reporting false. Pass reportingOnly=true to keep only fillable rows. Large disclosures (for example E1-6) often contain dozens of application_requirement rows and unanswered points. Agents should pass hasContent=true and excludeDatapointTypes: ["application_requirement"] (or datapointType: ["standard", "general"]) so one call returns only answered, non-AR datapoints. Both filters are string arrays — a single type is still ["standard"], not a bare string. Use maxLevel to keep parent/summary rows (datapointLevel <= maxLevel). Use limit/offset to page; when limit is set the response includes total, count, offset and hasMore. Omitting filters and limit returns every datapoint, which can blow the context window. counts.progress is the CMS progress denominator (disclosure.totalDatapoints, typically non-AR rows used for completion). counts.all is every active datapoint in the disclosure including application_requirement. counts.filtered is after the filters on this call. counts.returned / datapointCount is the length of the datapoints array in this response. description is plain text; descriptionHtml is only present when the CMS source contained markup. eurLexUrl is the legal-text deep link when the CMS has base_url/doc_id.

cards_report_get_disclosure_summary

ChatGPT
Returns the current summary text for a disclosure in a report. Generated summaries are not stored separately; after the user accepts one it is written to the disclosure's general datapoint. This tool returns that datapoint's current text when present. It never triggers generation.

cards_report_get_status

ChatGPT
Returns just the workflow status of a report, for example 'active', 'finalized' or 'archived'.

cards_report_get_topic

ChatGPT
Returns one topic inside a report, including its guidance text.

cards_report_list

ChatGPT
Lists the sustainability reports in the current organization that this token may read. If the token is restricted to selected reports, only those appear. Start here to find a report ID for the other report tools.

cards_report_list_changes

ChatGPT
Returns datapoints in a report whose stored answer was updated at or after since. Use this for questions such as 'what changed this week?' Each row is the current value with who last edited it and when. Datapoints turned off (active is false) are omitted. Pass includePrevious=true to also attach the immediately previous snapshot (content, actor, versionId) so you can summarise what changed without one history call per datapoint. For the full revision list of one datapoint, use cards_report_get_datapoint_history. Call cards_report_get_datapoint for the full current answer payload.

cards_report_list_copy_items

ChatGPT
Lists fields in a copied report that still need a mapping decision (or already have one, if open=false). Each item is one target datapoint: the previous report's answer (source), any proposed value, the new field's type/shape, and why it was not copied automatically (stage: clean/parsed/ai/manual; note). Walk these with your own reasoning — accept with cards_report_accept_copy_item or skip with cards_report_skip_copy_item. Paginate with limit/offset; follow hasMore. Requires mcp:read:reports. The report must have been created by a copy.

cards_report_list_copy_targets

ChatGPT
Lists the ESRS versions a report can be copied into: its own version (a roll-forward into a new period) plus any newer standard that has an active datapoint mapping. Use this before cards_report_copy. Other pairs (for example ESRS to VSME) are refused. Requires mcp:read:reports.

cards_report_list_disclosures

ChatGPT
Lists the disclosure requirements under one topic of a report, including materiality and completion progress.

cards_report_list_threads

ChatGPT
Returns comment threads on datapoints in a report, each with comments in created-at order (author, text, timestamps), a resolved flag, status, and which datapoint/disclosure/topic the thread belongs to. Filter with disclosureId (for example 'E1-1') and/or datapointId. unresolvedOnly defaults to false; pass true to keep only active, unresolved threads. Use this to answer what is blocked and why — unresolved threads are the discussions behind open questions, missing data, and work waiting on someone else. To add a comment or resolve a thread, use cards_report_create_comment and cards_report_resolve_thread (mcp:write:comments).

cards_report_list_topics

ChatGPT
Lists the ESRS topics in a report together with their completion progress. Use it to see how far along a report is, topic by topic.

cards_report_resolve_thread

ChatGPT
Marks a comment thread as resolved. Pass comment with a short closing note — recommended, because the thread then drops out of unresolvedOnly listings and otherwise nobody sees why it closed. Not required. The resolve is stored as the token's user. Requires mcp:write:comments and write access to an active report. Does not delete the thread or its comments.

cards_report_restore_datapoint

ChatGPT
Writes a previous history snapshot back onto the live datapoint. This is a new write — intermediate versions are not deleted, and history grows. versionId comes from cards_report_get_datapoint_history (history rows only; current has no versionId until it is overwritten). Restores the full snapshot: content, completed, active and contentOrigin. The new current row has restoredFromVersionId set so history can show it was a restore. actor.type is mcp_token; lastUpdatedBy is the human whose permission was used. Agents may restore any version they can read, including human writes — same rule as overwrite. Inactive datapoints are refused — turn them on with cards_report_set_datapoint_active first. Requires mcp:write:reports and write access to an active report.

cards_report_set_datapoint

ChatGPT
Writes the recorded answer for one fillable datapoint in an active report. Overwrites the current value (the previous version is archived to history). Fill in the write example that cards_report_get_datapoint returns (write.argument, write.example, write.notes): value for text/number/boolean/date/enum, fields for object datapoints, rows for tables. Text is plain text or markdown; HTML is refused. Numeric CMS fields are stored as JSON numbers, not strings. contentOrigin is set to ai; actor.type is mcp_token. Requires mcp:write:reports and write access. reporting must be true (label-only rows such as E1-6-44-a cannot be written). Inactive datapoints (active is false) are refused — turn them on with cards_report_set_datapoint_active first. Pass ifEmpty=true to refuse when a value is already stored. completed is optional and defaults to leaving the current flag unchanged.

cards_report_set_datapoint_active

ChatGPT
Turns one datapoint on or off in a report without changing its stored answer. Off (active=false) removes it from the report, including exports, publications and MCP reads — the same setting as Active datapoint in the app. The previous flag is archived in history with actor.type mcp_token. Requires mcp:write:reports and write access to an active report. This is the only write that is allowed while the datapoint is off.

cards_report_set_datapoint_completed

ChatGPT
Marks one datapoint as completed or not completed without changing its stored answer. This updates report progress counts. The previous completed flag is archived in history with actor.type mcp_token. Requires mcp:write:reports and write access to an active report. Inactive datapoints are refused — turn them on with cards_report_set_datapoint_active first. To also change the answer, use cards_report_set_datapoint.

cards_report_set_disclosure_material

ChatGPT
Marks a disclosure in a report as material or not material, the same switch as in the app. Mandatory disclosures (isMandatory in cards_report_list_disclosures) are always material and cannot be set to false. Requires mcp:write:reports and write access to an active report.

cards_report_set_responsible_user

ChatGPT
Sets or clears who is responsible for a report, topic, disclosure, or datapoint. Responsibility is stored on the report, topic or disclosure row — a datapoint assignment is written on its parent disclosure (the same place the app stores it). Pass userId to assign, or null to clear. comment is encouraged when assigning: it is posted on the datapoint when one is given, and the assignee is @-mentioned so they get the same notification as in the app. Requires mcp:write:reports and write access to an active report.

cards_report_skip_copy_item

ChatGPT
Dismisses one mapping queue item without writing an answer into the report. The previous value stays in the item's source for later; the target field is left empty. Requires mcp:write:reports and write access to an active report.

cards_report_update

ChatGPT
Changes a report's name and/or reporting period. Pass only the fields to change; the rest are kept. The ESRS version and language cannot be changed here. Requires mcp:write:reports and write access to an active report.

cards_resources_list

ChatGPT
Lists the curated external references shown on the product home page: GHG Protocol standards and guidance, the ESRS legal text in EUR-Lex, and EFRAG's datapoint Excel. Filter by language (en or sv). Official EUR-Lex texts and per-topic/disclosure deep links are also MCP resources (resources/list, resources/read) — those return the official URL plus CMS guidance, not the Official Journal HTML. These catalogue rows are global references, not organization files — use cards_organization_files_list for uploaded library files.

cards_search_organization_files

ChatGPT
Vector (semantic) search over the indexed contents of the organization's files. Matches on meaning rather than keywords, so a full question or a description of the evidence needed works better than search terms. Returns the matching passages with their source file, ranked by similarity. Files with isIndexed: false will not appear here; use cards_organization_files_get_content to read them in full. Use this to find supporting evidence for a disclosure or datapoint.

cards_search_reports

ChatGPT
Full-text search across the answers recorded in the organization's active reports, and across topic, disclosure and datapoint identifiers and names. Pass a reportId to focus on one report. For a known datapoint ID across every report (for example scope 1 in E1-6-44-a), use cards_report_get_datapoint_across_reports instead — it is faster and returns the stored values. To browse the standard itself without report context, use cards_esrs_search.

cards_user_get_recent_work

ChatGPT
Returns the disclosures the current user last worked on — the same 'continue your work' list shown on the home page and in each report. Always the token's user, never another person. Pass a reportId to limit to one report; omit it to look across every active report in the organization.

cards_user_list_assignments

ChatGPT
Returns reports, topics and disclosures in the organization where the current user is set as responsible_user. Always the token's user. Use this to answer what is waiting on me; use cards_user_get_recent_work for what I last edited.

cards_workflow_get

ChatGPT
Returns one assisted workflow template with its sections and the steps inside them, in the order they are presented. Returns the definition only; call cards_workflow_report_get for the answers a run produced.

cards_workflow_list

ChatGPT
Lists the assisted workflow templates available to run. These are global definitions of a guided, AI-assisted process, not organization content.

cards_workflow_report_get

ChatGPT
Returns one assisted workflow run: its progress plus every section and step, and whether each step has been answered. Answers are omitted by default because they can be long — set includeAnswers to true to get the generated content as well. A workflow report is one run of an assisted workflow. It is not a sustainability report.

cards_workflow_report_get_step

ChatGPT
Returns the generated answer for a single step of an assisted workflow run. Use this instead of includeAnswers when you only need one step. A workflow report is one run of an assisted workflow. It is not a sustainability report.

cards_workflow_report_list

ChatGPT
Lists the assisted workflow runs belonging to the current organization, with their status and how far through the steps each one is. A workflow report is one run of an assisted workflow. It is not a sustainability report.

Capabilities

Writes

Example Prompts

Click any prompt to copy it.

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/the-cards"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/the-cards)