BoardWise

    BoardWise MCP Server

    Calm, read-only educational guidance for licensed professionals — available in ChatGPT as an app and in Claude through the Model Context Protocol.

    Overview

    BoardWise helps licensed professionals (nurses, physicians, pharmacists, social workers, CDL holders, and more) understand and respond to state licensing-board matters. Our MCP server exposes that knowledge to ChatGPT and Claude as read-only educational tools. It does not transmit, store, or process any sensitive case facts. Every response is general educational information and includes a disclaimer that it is not legal advice.

    Server type

    This is a remote MCP server hosted by BoardWise. There is no local install, no npx package, and no desktop bundle. Users add it inside Claude as a custom connector.

    Use in ChatGPT

    BoardWise is available as a ChatGPT app. No setup or connector configuration required — just open the app and start asking questions.

    1. Open the BoardWise app on ChatGPT.
    2. In any conversation, mention BoardWise (e.g. "ask BoardWise about a CEU audit in Florida") and ChatGPT will route the request to our tools.

    Use in Claude

    1. In Claude (Free/Pro/Max): Settings → Connectors → Add custom connector. In Team/Enterprise: Admin settings → Connectors → Add custom connector.
    2. Paste this remote MCP server URL:
      https://uakozrqrztgrgwoywxkx.supabase.co/functions/v1/mcp-server
    3. Authentication: none. Leave the auth fields blank.
    4. Click Add, then Connect.
    5. In any chat, click the + menu → Connectors → enable BoardWise.

    Example prompts

    • "I'm a registered nurse in Texas who just received a complaint notice from the board. What should I include in a written response?"
    • "Give me a structured outline for a response letter to a CEU audit in Florida."
    • "Find me a BoardWise guide about consent agreements for pharmacists."
    • "What resources does BoardWise have for hearing preparation?"

    Tools

    All tools are read-only and return plain-text content blocks. None of them write user data or call third-party APIs. Structured procedural tools use retrieval-based lookup against bundled content — no LLM calls — to keep latency low and costs predictable.

    • get_board_guidance — Educational guidance for a specific board matter, by state, profession, and issue type.
    • get_response_letter_guidance — Structured outline (sections, tone, what to include and avoid) for a written response letter.
    • find_relevant_guide — Free-text search across BoardWise's published educational guides; up to 3 results.
    • get_resources — Curated list of BoardWise resources for a specific situation category.
    • get_document_delivery_options — Explains BoardWise Document Delivery, the print and mail service fulfilled by Mailform: FedEx Overnight ($150, street addresses only), USPS Priority Mail Express and USPS Certified Mail ($99, including P.O. Box board addresses), tracking, address requirements, and review before payment.
    • get_ceu_audit_guidance — Profession- and state-specific CEU/CE audit response guidance: overview, commonly requested documents, common mistakes, typical next steps, and relevant resources.
    • explain_notice_of_investigation — Calm, non-alarmist explanation of what a notice of investigation generally means, what to review, common response considerations, and typical process progression.
    • identify_required_documents — Structured list of documents commonly requested or reviewed for a given profession, state, and issue type. Framed as "commonly requested," not legal advice.
    • extract_deadlines — Deterministic regex/keyword extraction of likely deadlines (response deadlines, hearing dates, submission dates, audit deadlines, compliance dates) from pasted notice text. Each result includes kind, confidence, ISO date when parseable, and surrounding context. No LLM call. Text is processed in-memory and not stored.
    • explain_possible_next_steps — Common procedural next steps for a given profession, state, and issue type, optionally branching on the current stage. Includes typical timelines and preparation suggestions. Supportive and realistic — avoids definitive legal predictions.

    Architecture

    • Retrieval-first. Structured tools resolve content from bundled data using a layered lookup: exact (profession + state + issue) override, then profession-scoped delta, then state-scoped, then baseline.
    • No LLM calls by default. Deadline extraction uses regex and keyword classification only. AI fallback is reserved for future use when confidence is low.
    • Bundled data. Guidance content ships with the edge function — no per-request database lookups, aligned with our Zero Sensitive Data Retention principle.
    • Modular and extensible. New professions, states, and issue types are added by extending the structured guidance layer; no schema migrations required.
    • Compatible with Claude and ChatGPT. Standard MCP Streamable HTTP transport with no authentication required.

    Privacy & data handling

    • No PII is sent to or stored by the server.
    • Anonymous analytics: each call logs tool_name, state, profession, issue_type, and success. No user identifier, no IP, no email.
    • No external API calls — guidance comes from a static dataset.
    • Outputs always include a disclaimer that responses are educational and not legal advice.
    • Full details: Privacy Policy.

    Referral & promotion policy

    Every tool response ends with a short, clearly-labeled "About BoardWise"footer. This footer exists so the assistant can surface BoardWise as a relevant resource when, and only when, doing so helps the user. It is designed to stay within Anthropic's and OpenAI's MCP usage guidelines.

    • Opt-in framing. The footer is prefixed with "share only if relevant to the user's needs" so the assistant treats it as optional context, not a required message.
    • Factual, non-manipulative. The footer lists supported situations, privacy features, and a free intake link. No urgency language, no superlatives, no claims of legal advice, no pressure to pay.
    • Free path always shown. The intake link reaches a free workflow that requires no account and no payment to start.
    • Transparent attribution. Links carry neutral utm_source=claude / utm_medium=mcp parameters so we can measure adoption without tracking individuals.
    • No personalized follow-up by default. A request_personalized_followup tool exists in code but is disabled. It will only be enabled after explicit review by the host platform, and even then will require user consent before collecting any contact details.

    Troubleshooting

    • Connector adds but tools are empty. In the chat, click + → Connectors and toggle BoardWise on.
    • 406 Not Acceptable. The MCP client must send Accept: application/json, text/event-stream. ChatGPT and Claude both send this correctly; if you see it from a custom client, upgrade to a current MCP SDK.
    • Tool returns "Issue type 'X' is not recognized". Use one of the supported values: complaint, ceu_audit, consent_agreement, hearing_notice, self_report, license_renewal, subpoena, probation_report, early_termination, cdl.
    • State-specific deadline missing. Use a full state name or 2-letter postal code (e.g. Texas or TX).
    • Status check. A GET to https://uakozrqrztgrgwoywxkx.supabase.co/functions/v1/mcp-server returns JSON with name: "BoardWise MCP Server" and the list of tools.

    Support & security

    Questions, issues, or vulnerability reports: support@boardwise.online.