MCP App Store
by KeenEthics
Genny – Generative Process and Decision Modeler by Blue Polaris icon

Genny – Generative Process and Decision Modeler by Blue Polaris MCP Server for Claude

By Blue Polaris Inc

Overview

Genny, by Blue Polaris brings AI-powered decision and process modeling directly into Claude. Give Claude your existing business rules, requirements, policies, code, or spreadsheets, and Genny transforms that logic into a visual, standards-based Decision Model & Notation (DMN) decision model or a Business Process BPMN —ready to review, refine, and edit in the Blue Polaris Modeler. Instead of manually translating complex or legacy requirements, Genny accelerates the path to models, helping teams understand how models work, uncover complexity, and create a standardized foundation for automation and modernization. Genny can work across a wide range of source materials, including DRL, Java, COBOL, PDF, Word, Markdown, text, CSV, Excel, JSON, and XML. Multiple files can be analyzed together and converted into a unified DMN and BPMN model, while larger documents and rule sets can be processed asynchronously without interrupting your Claude workflow. Once generated, Genny provides a direct link to open the model in the Blue Polaris Modeler, where teams can visualize, review, and evolve their decision logic and deploy for execution.

How to connect

  1. Claude: Open the Genny – Generative Process and Decision Modeler by Blue Polaris listing in the Claude connectors directory and choose Connect. Claude listing Genny – Generative Process and Decision Modeler by Blue Polaris exposes 14 tools here.

Tools

check_dmn_conversion

Claude
Returns the current status of a job started by start_dmn_conversion. Response statuses: running — job still processing; elapsed_s reports time so far. success — diagram_url and complete_model_id are present. error — conversion failed; the message field contains the reason. Pass verbose=True to include status_history and timings_s.

connect_dfm

Claude
Confirms the active Bluepolaris connection is ready to use. Response statuses: success — connection is ready; tools can be called. Returns account_id, user_id, and api_url. auth_required — Clerk sign-in has not completed (or the session expired); the response includes an auth_url — open it, sign in, then retry. trial_expired — the account's free-trial window has ended; the response includes a contact_url. Typical first-time use: 1. User adds the connector and asks Claude to connect to DFM (this tool runs). 2. If auth_required, the user opens auth_url, signs in via Clerk, and returns to Claude. 3. User retries; connection is ready for all subsequent tool calls.

convert_to_bpmn

Claude
Synchronously converts one or more uploaded files (an SOP, PDF, DOC, etc.) into a BPMN process diagram. Same upload + conversion pipeline as convert_to_dmn, just requesting the Bluepolaris DFM platform's outputType=process rendering instead of dmn. Returns a diagram_url plus complete_model_id on success. Parameters: upload_id — a single upload_id (str) or a list for multi-file jobs. Each must come from a successful create_upload_ticket + PUT upload cycle. target_language — language for generated artifacts (default English). timeout — REST upload timeout in seconds. event_timeout — maximum seconds to wait for the CompleteModel event. verbose — include status_history and timings_s in the response. model — "Production" or "Genny". Optional: if omitted, PDF-only input defaults to "Genny"; anything else defaults to "Production" (the only model non-PDF inputs support). The resolved value is echoed back in the response's model field. .zip inputs are rejected.

convert_to_dmn

Claude
Synchronously converts one or more uploaded files into a DMN decision diagram. Returns a diagram_url plus complete_model_id on success. The companion tool start_dmn_conversion performs the same operation asynchronously and is appropriate when the conversion may exceed the synchronous request window. Parameters: upload_id — a single upload_id (str) or a list for multi-file jobs. Each must come from a successful create_upload_ticket + PUT upload cycle. output_type — "dmn" or "process" (default "dmn"). target_language — language for generated artifacts (default English). timeout — REST upload timeout in seconds. event_timeout — maximum seconds to wait for the CompleteModel event. verbose — include status_history and timings_s in the response. model — "Production" or "Genny". Optional: if omitted, PDF-only input defaults to "Genny"; anything else defaults to "Production" (the only model non-PDF inputs support). The resolved value is echoed back in the response's model field. .zip inputs are rejected.

create_upload_ticket

Claude
Reserves a one-time upload slot on the server for a single file. Returns an upload_id, an upload_url, and an upload_token. The client uploads the raw file bytes via: PUT <upload_url> X-Upload-Token: <upload_token> The resulting upload_id is consumed by the conversion tools (convert_to_dmn, import_file, import_files, start_dmn_conversion). Tickets expire after a fixed TTL.

dfm_build_drd

Claude
Adds decisions, input data, knowledge sources, and requirement arrows to a DRD. Everything is addressed by name — a node that already exists on the diagram is reused rather than duplicated. Positions are worked out automatically; do not try to supply coordinates (use dfm_move_node afterwards if the user wants a specific placement). Send the whole picture in one call where you can — it lays out better than several small calls. Parameters: diagram_id — the diagram to build in. decisions — names of decisions, e.g. ["Candidate Meets Requirements"]. input_data — names of input data, e.g. ["Years Of Experience"]. knowledge_sources — names of policies, regulations, or analytics that give a decision its authority, e.g. ["Screening Policy"]. requirements — the arrows, as [{"decision": "<decision name>", "requires": "<name of the decision, input data, or knowledge source it needs>"}]. A decision requires; nothing else does.

dfm_build_process

Claude
Adds lanes, tasks, gateways, events, and sequence flows to a BPMN process diagram. Mirrors dfm_build_drd, but for process diagrams — see that tool's docstring for the general shape. diagram_id is a process id from dfm_find_process, not dfm_find_diagram. CONFIRMED 2026-09-22 both by schema inspection and a live round-trip (added a lane + task + end event + sequence flow to "Procurement Process", verified every id showed up in the right places, then removed all four and verified the diagram was back to its original 27 nodes / 26 connectors). A lane's flowNodeRefs list is what actually assigns a task/gateway/event to it — checked exhaustively across all 3 process diagrams on this account: every non-lane node belongs to exactly one lane, no gaps, no duplicates. Position is NOT something to set here — every node across all 3 diagrams had a "0 0" stored location despite each rendering a real layout, so the modeler computes layout client-side from lane membership and the sequence-flow graph; new nodes are written the same way ("0 0") and lay out automatically once flowNodeRefs and sequence_flows are right. Everything is addressed by name — reusing an existing element rather than duplicating it. Parameters: diagram_id — the process diagram to build in. lanes — lane names to ensure exist, e.g. ["Finance Department"]. tasks — [{"name": "...", "lane": "<lane name>"}]. "lane" is optional but the element won't be assigned to any swimlane without it. The lane must already be on the diagram or be listed in lanes in this same call — a node naming any other lane is skipped, not created. gateways — same shape as tasks, e.g. [{"name": "Approved?", "lane": "Procurement Manager"}]. events — [{"name": "...", "type": "start"|"end", "lane": "..."}]. "type" is required — there's no way to guess which one a bare event name means. sequence_flows — the arrows, as [{"from": "<node name>", "to": "<node name>", "condition": "<optional label, e.g. Yes/No>"}].

dfm_describe_drd

Claude
Reads what is currently on a Decision Requirements Diagram. Returns its decisions, input data, knowledge sources, and every requirement arrow actually drawn between them. Call this before dfm_build_drd, dfm_rename_node, or dfm_move_node so you can confirm with the user what's already there and which node/arrow they mean. Parameters: diagram_id — the diagram's id, from dfm_find_diagram or the user.

dfm_describe_process

Claude
Reads what is currently on a BPMN process diagram. Returns its lanes, tasks, gateways, events, and sequence flows, the same way dfm_describe_drd does for a DRD, plus created_date, created_by and source_files (see dfm_find_diagram). Call this before making changes to a process diagram so you can confirm with the user what's already there. CONFIRMED against a real process diagram: diagram_id here must be a process id from dfm_find_process, NOT a diagram_id from dfm_find_diagram — process diagrams live under a completely separate REST resource (processes/{id}) from DRD/ERD diagrams (diagrams/{id}); the two id spaces don't overlap, and this will error cleanly if given the wrong kind of id. See _DFM_PROCESS_LANE_TYPES above for how elements are classified. Parameters: diagram_id — the process diagram's id, from dfm_find_process or the user.

dfm_find_diagram

Claude
Finds a Decision Requirements Diagram (DRD) by (partial) name. Use this first when the user names a diagram rather than giving its id — show them the matches so they can confirm which one they mean before calling dfm_describe_drd, dfm_build_drd, dfm_rename_node, or dfm_move_node. Parameters: name — part of the diagram's name; empty lists the first 20.

dfm_find_process

Claude
Finds a BPMN process diagram by (partial) name. CONFIRMED: process diagrams live under a completely separate REST resource ("processes") from DRD/ERD diagrams ("diagrams") — a diagram_id from dfm_find_diagram will never match here, and vice versa. Use this first when the user names a process diagram rather than giving its id, then pass the returned id to dfm_describe_process. Parameters: name — part of the process diagram's name, or its full id; empty lists all processes. page — which page of 20 results to return, starting at 1 (see dfm_find_diagram). Each result also carries created_date, created_by and source_files, as described for dfm_find_diagram.

dfm_move_node

Claude
Repositions one node on a DRD, dragging its arrows along with it. Only use this when the user asks for a specific placement — layout is otherwise left exactly as a person arranged it. The canvas runs x to the right and y DOWNWARD, so a smaller y is higher up. Call dfm_describe_drd first if you need to know where things currently are. Parameters: diagram_id — the diagram the node is on. name — the node to move (as shown by dfm_describe_drd). x — horizontal position, increasing to the right. y — vertical position, increasing downward.

dfm_rename_node

Claude
Renames a decision, input data, or knowledge source. The rename applies to the object across DFM, not just this diagram — check with the user if the name might be reused elsewhere before calling this. Parameters: diagram_id — the diagram the node is on. name — its current name (as shown by dfm_describe_drd). new_name — what to call it instead.

dfm_rename_process

Claude
Renames a task, gateway, event, or lane on a BPMN process diagram. Mirrors dfm_rename_node, but for process diagrams. Call dfm_describe_process first so you can confirm with the user what's already there and which element they mean. diagram_id is a process id from dfm_find_process, not dfm_find_diagram — see dfm_describe_process. CONFIRMED 2026-09-22 by a live round-trip against the "Procurement Process" diagram (rename, verify via re-fetch, rename back) — PATCH returned 200 and the new name was actually persisted, not just accepted. Unlike DRD, a goNodes entry isn't its own top-level REST object — it's one entry in a dict inside processImplementation, itself a JSON string on the outer process resource. This PATCHes back the ENTIRE process object as fetched, with only the target node's name changed inside the re-serialized processImplementation string. Parameters: diagram_id — the process diagram's id, from dfm_find_process. name — its current name (as shown by dfm_describe_process). new_name — what to call it instead.

generate_request_id

Claude
Returns a freshly generated UUID4 string in the uuid field. Utility for constructing requestid or transactionid values when calling the DFM API directly. The conversion tools generate their own identifiers internally and do not need this.

get_identity

Claude
Returns the decoded identity associated with the active DFM auth token. Response fields: account_id — DFM account id decoded from the JWT. user_id — DFM user id decoded from the JWT. api_url, ws_url, modeler_url — the configured DFM endpoints. The raw token is never returned. Useful for verifying that authentication is in place before running a conversion.

import_file

Claude
Synchronously converts a single uploaded file into a DMN diagram. Accepts exactly one upload_id and returns a diagram_url. For inputs whose conversion may exceed the synchronous request window, use start_dmn_conversion instead. Parameters: upload_id — upload_id from a successful create_upload_ticket + PUT. target_language — language for generated artifacts (default English). output_type — "dmn" or "process" (default "dmn"). timeout — REST upload timeout in seconds. event_timeout — maximum seconds to wait for the CompleteModel event. verbose — include status_history and timings_s in the response. model — "Production" or "Genny". Optional: if omitted, a PDF file defaults to "Genny"; anything else defaults to "Production" (the only model non-PDF inputs support). The resolved value is echoed back in the response's model field. .zip inputs are rejected.

import_files

Claude
Synchronously converts a set of related uploaded files into a single DMN diagram. Intended for sources split across multiple files (for example a .drl plus its imported .java classes, or a set of CSVs). Returns one diagram_url covering all inputs. Parameters: upload_id — list of upload_ids (or a single str) from successful create_upload_ticket + PUT cycles. target_language — language for generated artifacts (default English). output_type — "dmn" or "process" (default "dmn"). timeout — REST upload timeout in seconds. event_timeout — maximum seconds to wait for the CompleteModel event. verbose — include status_history and timings_s in the response. model — "Production" or "Genny". Optional: if omitted, PDF-only input defaults to "Genny"; anything else (including a mix of PDF and non-PDF files) defaults to "Production", the only model non-PDF inputs support. The resolved value is echoed back in the response's model field. .zip inputs are rejected.

start_dmn_conversion

Claude
Starts an asynchronous DMN-conversion job and returns a job_id immediately. Intended for inputs whose conversion may exceed a synchronous request window (PDFs, large rule sets). The companion tool check_dmn_conversion polls the resulting job_id until the diagram is ready. Parameters: upload_id — a single upload_id (str) or a list for multi-file jobs. Each must come from a successful create_upload_ticket + PUT upload cycle. model — "Production" or "Genny". Optional: if omitted, PDF-only input defaults to "Genny"; anything else defaults to "Production" (the only model non-PDF inputs support). output_type — "dmn" or "process" (default "dmn"). target_language — language for generated artifacts (default English). timeout — REST upload timeout in seconds. event_timeout — maximum seconds to wait for the CompleteModel event. .zip inputs are rejected.

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/genny-generative-process-bpmn-decision-dmn-modeler"><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/genny-generative-process-bpmn-decision-dmn-modeler)