Documentation

What is MCP?

MCP — the Model Context Protocol — is an open standard created by Anthropic that allows AI assistants to connect to external tools and data sources. It works similarly to how a web browser connects to websites: your AI assistant sends a request, the MCP server returns data, and the AI incorporates that data into its response.

PolicyHQ is an MCP server. When you connect it to your AI assistant, you are giving that assistant permission to query our database of UK case law and legislation on your behalf. You do not need to type any special commands — the AI decides when to use PolicyHQ based on what you ask it.

MCP is now supported by Anthropic (Claude), OpenAI (ChatGPT), and Microsoft (Copilot). The setup process differs slightly for each platform, but the endpoint and authentication are the same.

Connecting PolicyHQ to Claude

These instructions apply to Claude.ai and Claude for Teams. For Claude API integration, see the tool reference section.

Open Claude.ai and go to Settings (your profile icon, bottom left).
Select Integrations from the settings menu.
Click Add MCP server.
Enter the PolicyHQ server URL: https://mcp.policyhq.co.uk/
Enter your API key when prompted. Your API key is available in your PolicyHQ dashboard.
Click Connect. PolicyHQ will appear in your integrations list.

Once connected, Claude will automatically use PolicyHQ when you ask legal questions that require UK case law or statute authority. You do not need to prefix your queries with any commands.

To verify the connection is working: ask Claude "What is the leading case on the duty of care in negligence?" Claude should respond with a cited answer and indicate that it used PolicyHQ.

Connecting PolicyHQ to ChatGPT

MCP support in ChatGPT is currently available in Developer Mode. Standard ChatGPT Plus users may need to wait for the general release of MCP support.

Open ChatGPT and navigate to Settings > Beta features.
Enable Model Context Protocol if shown as a beta option.
Go to Settings > Connected tools.
Click Add tool and select MCP server.
Enter the server URL: https://mcp.policyhq.co.uk/
Enter your PolicyHQ API key.
Save the connection.

If MCP integration is not yet available in your ChatGPT account, contact us at hello@policyhq.co.uk — we can advise on alternative connection methods as they become available.

Connecting PolicyHQ to Microsoft Copilot

For Microsoft Copilot in Microsoft 365, integration is managed by your organisation's IT administrator via Copilot Studio. PolicyHQ cannot be added by individual users without administrator access.

If you are an IT administrator:

Open Copilot Studio in your Microsoft 365 admin portal.
Navigate to Connectors > Custom connectors.
Create a new MCP connector with the endpoint https://mcp.policyhq.co.uk/.
Configure authentication using your organisation's PolicyHQ API key (Enterprise plan required for multiple users).
Assign the connector to the relevant Copilot agent or team.

For Enterprise deployment support, contact us at hello@policyhq.co.uk. We provide dedicated onboarding assistance for firm-wide Copilot integrations.

Searching case law and legislation

Once connected, you can ask your AI assistant legal questions in natural language. PolicyHQ's search tool — search_policy — uses semantic search to find relevant text by meaning, not just keyword matching.

Examples of what you can ask:

"What is the test for dishonesty in a criminal context?"
"What remedies are available to an employee dismissed during a probationary period?"
"Does the doctrine of promissory estoppel require detrimental reliance?"
"Find the relevant provisions of the Landlord and Tenant Act 1954 on business tenancy renewal."

PolicyHQ returns the exact paragraphs from real judgments or statute sections that answer the question. Your AI assistant then incorporates these into its response with full neutral citations.

Limitations: PolicyHQ covers UK case law from 2003 onwards and UK legislation currently in force. Older case law, unreported decisions, and non-UK authorities are outside our current coverage.

Verifying citations

The verification tool — verify_document — checks every legal citation in a document you have drafted. To use it:

Ask your AI assistant to draft or review a document containing legal citations.
Before finalising, ask the AI: "Please verify all the citations in this document using PolicyHQ."
The AI will call verify_document and return a report showing the status of each citation.
Review the report. Investigate any citations marked as uncertain or not found before filing.

Understanding the verification statuses:

Verified — the citation was found in our database and the associated content appeared consistent with the source text. This is a positive indication, not a guarantee of legal correctness.
Uncertain — the citation exists in our database, but the specific proposition claimed in your document could not be fully confirmed from the available text. Independent verification is advisable.
Not found — the citation does not appear in our database. It may be hallucinated, mis-cited, or outside our coverage. Do not use this citation without manual verification against the original source.
Outside coverage — the reference type is not in our corpus (for example, pre-2003 case law, non-UK sources, or unreported decisions). PolicyHQ cannot verify or refute these references.

Verification is an automated first-pass check. It is not a substitute for professional legal judgement. Solicitors remain personally responsible for the accuracy of all citations submitted to courts and tribunals.

Compliance dashboard

Your PolicyHQ dashboard is available at app.policyhq.co.uk. It provides:

A full log of every search query made via your API key(s), with timestamps and results.
A record of every verification report, with per-citation status and source references.
Usage analytics: queries per day, week, and month.
API key management: create, label, and revoke keys for team members or specific AI platforms.
CSV export of query logs and verification reports for your compliance files.
Email alerts when not-found citations are detected in verifications.

The dashboard is designed to support your obligations under the SRA Standards and Regulations regarding appropriate governance and supervision of technology used in legal practice.

Tool reference: search_policy

Semantic search across UK case law and legislation.

Parameters:
  query (string, required)
    Natural language search query.

  source (string, optional, default: "all")
    "case_law" | "legislation" | "all"

  court (string, optional)
    Filter by court abbreviation.
    Values: UKSC, UKHL, EWCA, EWHC, UKUT, UKFTT, EAT

  year_from (integer, optional)
    Start year. Minimum 2003.

  year_to (integer, optional)
    End year.

  limit (integer, optional, default: 10, max: 50)
    Maximum results to return.

Returns:
  Array of text chunks, each containing:
  - citation (neutral citation string)
  - court, date
  - paragraph number (for judgments)
  - source_url (link to primary source)
  - text (the relevant passage)

Tool reference: get_judgment

Retrieve the full text of a judgment by neutral citation.

Parameters:
  citation (string, required)
    UK neutral citation. Example: "[2024] UKSC 1"

Returns:
  Full judgment text with numbered paragraphs, metadata
  (court, date, parties, judge(s)), and source URL.

Tool reference: get_legislation

Retrieve a specific section of an Act of Parliament or Statutory Instrument.

Parameters:
  act (string, required)
    Full Act title. Example: "Employment Rights Act 1996"

  section (string, required)
    Section reference. Example: "s.92" or "section 92"

Returns:
  Section text, amendment status, and link to legislation.gov.uk.

Tool reference: verify_document

Verify all legal citations in a document.

Parameters:
  document_text (string, required)
    Full text of the document. Maximum 50,000 characters.
    Citations are extracted automatically.

  check_content (boolean, optional, default: true)
    Whether to check content claims against source text,
    not just citation existence.

Returns:
  Array of per-citation results, each containing:
  - citation_text (the extracted citation string)
  - status: "verified" | "uncertain" | "not_found" | "outside_coverage"
  - notes (explanation of the status)
  - source_url (if found)

Tool reference: verify_citations

Verify a specific list of citations.

Parameters:
  citations (array, required)
    Array of objects, each containing:
    - citation_text (string, required): the citation string
    - claim_text (string, optional): the specific claim made

Returns:
  Per-citation verification results (same as verify_document).

Tool reference: check_status

Check whether a case has been overruled or a statute is in force.

Parameters:
  reference (string, required)
    A neutral citation or legislation reference.
    Example: "[2003] UKHL 47" or "Occupiers' Liability Act 1984 s.1"

Returns:
  Status: "in_force" | "repealed" | "amended" | "overruled" | "unknown"
  With relevant dates and links to primary source.

SRA obligations and PolicyHQ

The Solicitors Regulation Authority has made clear that solicitors using AI tools in legal practice remain personally responsible for the work they submit. The SRA Standards and Regulations require firms to have appropriate systems and controls when using technology, including AI.

PolicyHQ supports compliance in two ways:

Citation verification — providing an automated first-pass check on AI-generated legal references before documents are finalised.
Audit trail — the compliance dashboard provides a complete log of every query and verification, with timestamps, which can be retained as evidence of appropriate governance.

PolicyHQ does not itself constitute compliance with SRA obligations. Solicitors must exercise professional judgement and independent verification of all legal references. The verification tool is a tool, not a substitute for a qualified lawyer's review.

Data coverage

Case law: The National Archives Find Case Law corpus. Coverage begins in 2003. Courts included: UKSC, UKHL, EWCA Civil Division, EWCA Criminal Division, EWHC (all divisions: Administrative, Business, Chancery, Commercial, Family, Intellectual Property, King's Bench, Patents, Planning, Technology and Construction), UKUT, UKFTT, EAT, and others.

Legislation: Acts of Parliament and Statutory Instruments from legislation.gov.uk, with amendment status where available. Scottish Parliament legislation, Welsh Senedd legislation, and Northern Ireland Assembly legislation are included where published on legislation.gov.uk.

Outside coverage: Pre-2003 case law, unreported decisions, tribunal decisions not published on The National Archives, foreign law, EU retained law not on legislation.gov.uk, regulatory guidance and codes (FCA, SRA, ICO, etc.), and academic commentary.

Frequently asked questions

Can I use PolicyHQ without an AI assistant?

PolicyHQ is designed as an MCP server — infrastructure for AI assistants. There is no standalone web interface for legal research. If you need direct API access for custom integration, contact us at hello@policyhq.co.uk.

How current is the database?

Case law: judgments are added from The National Archives as they are published, typically within 24–48 hours of publication. Legislation: updates are synchronised from legislation.gov.uk on a regular basis. There may be a short delay between a judgment or amendment being published and appearing in PolicyHQ.

What is the difference between "verified" and legally correct?

"Verified" means that the citation was found in our database and the associated content appeared consistent with the source text at the time of checking. It does not mean that the legal proposition is correctly stated, correctly applied, or that the case is still good law. Solicitors must exercise independent professional judgement.

Does PolicyHQ store my documents?

No. Document text submitted for verification is processed in memory and discarded after the verification report is generated. We do not store the full text of any document. We do retain the extracted citations and their verification status in your compliance dashboard query log. See our Privacy Policy for full details.

Can I share my API key with colleagues?

You should create separate API keys for each team member or AI platform using the dashboard. Sharing a single API key between users means their usage is not separately logged, which undermines the audit trail. Professional and Enterprise plans allow multiple API keys.