教程 07: MCP Integrations for Legal Workflows

Search for Delaware cases on limitation of liability clauses in
software agreements from the last 3 years.

Search CourtListener for Ninth Circuit cases discussing
software license enforceability.

npm install @modelcontextprotocol/sdk zod

// imanage-mcp.js
import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
import { z } from "zod";
import iManageClient from "imanage-api-client";
 
const server = new McpServer({ name: "imanage", version: "1.0.0" });
 
server.registerTool(
  "search_documents",
  {
    description: "Search iManage for documents",
    inputSchema: {
      query: z.string().describe("Search query"),
      workspace: z.string().describe("Workspace ID"),
    },
  },
  async ({ query, workspace }) => {
    const client = new iManageClient(process.env.IMANAGE_CREDENTIALS);
    const results = await client.search({ query, workspace });
    return { content: [{ type: "text", text: JSON.stringify(results) }] };
  },
);
 
server.registerTool(
  "get_document",
  {
    description: "Retrieve document content",
    inputSchema: {
      documentId: z.string().describe("Document ID"),
    },
  },
  async ({ documentId }) => {
    const client = new iManageClient(process.env.IMANAGE_CREDENTIALS);
    const document = await client.getDocument(documentId);
    return { content: [{ type: "text", text: JSON.stringify(document) }] };
  },
);
 
const transport = new StdioServerTransport();
await server.connect(transport);

I need to research the enforceability of liquidated damages
clauses in software agreements under Delaware law.

Workflow:
1. Search Midpage for Delaware cases on liquidated damages
2. Search CourtListener for federal cases applying Delaware law
3. Pull our firm's memo on liquidated damages from SharePoint
4. Synthesize into a research summary

Let me research this across multiple sources...

## Midpage Results (Delaware State Cases)
[Assistant queries Midpage, returns cases]

## CourtListener Results (Federal/Delaware Law)
[Assistant queries CourtListener, returns additional cases]

## Firm Precedent (SharePoint)
[Assistant retrieves firm memo]

## Synthesis
Based on my research across these sources:

The leading Delaware case is [X], which established...
Federal courts applying Delaware law have held...
Our firm's prior analysis in the [Client] matter noted...

**Key Takeaways**:
1. Delaware generally enforces liquidated damages if...
2. The "reasonableness" test requires...
3. Our recommended approach is...

**Citations** (verified via Midpage citator):
- [Citation 1] - Still good law
- [Citation 2] - Still good law

/review-contract
[Upload software agreement]

[Assistant identifies concerning liability provision under CA law]

"I've flagged the liability limitation as potentially problematic.
Let me research California's position on liability caps in
software agreements..."

[Assistant queries Midpage for CA cases]

"Based on my research, California courts have held that [X].
The clause as drafted may be enforceable because [Y], but
you should consider [Z] modification."

We're acquiring a company in the healthcare space.
Research the regulatory landscape:

1. Search our SharePoint for prior healthcare M&A memos
2. Search Midpage for recent HIPAA enforcement actions
3. Search for FTC healthcare antitrust cases
4. Generate a due diligence issue checklist

What MCP servers do you have access to right now?
List each server and its available tools.

Test the Midpage search tool by searching for "contract breach"
and report any errors.

Your Query
    ↓
Your assistant (vendor servers)
    ↓
MCP Server (Your Control)
    ↓
External Service (Midpage/iManage/etc.)

{
  "mcpServers": {
    "congress": {
      "command": "npx",
      "args": ["-y", "@open-legal-tools/congress-mcp"],
      "env": {
        "CONGRESS_API_KEY": "optional-api-key"
      }
    }
  }
}

Research task:
Track amendments to the California CCPA across its
legislative history and identify key changes by session.

Assistant multi-step workflow:
1. Query Congress API for CCPA bill history
2. Retrieve full text for versions from 2018, 2020, 2023
3. Compare key sections (scope, penalties, exemptions)
4. Generate side-by-side amendment analysis
5. Flag enforcement implications of changes

{
  "mcpServers": {
    "congress": {
      "command": "npx",
      "args": ["-y", "@open-legal-tools/congress-mcp"],
      "tools": [
        "search_bills",
        "get_bill_versions",
        "track_amendments",
        "retrieve_sponsor_info"
      ]
    },
    "regulationtracker": {
      "command": "npx",
      "args": ["-y", "regulation-tracker-mcp"],
      "env": {
        "FEDERAL_REGISTER_API": "api-key"
      }
    }
  }
}

Query: "Search all HIPAA regulations (45 CFR 160-164)
related to breach notification requirements and identify
which sections were modified in the last 3 years."

Assistant workflow:
1. Query regulatory database for HIPAA sections
2. Cross-reference Federal Register for amendments
3. Pull relevant case law on HIPAA enforcement
4. Compile regulatory timeline
5. Provide implementation checklist for compliance

// bill-tracking-mcp.js - Custom MCP for monitoring bills
import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
import { z } from "zod";
 
const server = new McpServer({ name: "bill-tracker", version: "1.0.0" });
 
server.registerTool(
  "monitor_bills",
  {
    description: "Set up alerts for bills matching criteria",
    inputSchema: {
      keywords: z.array(z.string()).describe("Bill topic keywords"),
      jurisdictions: z.array(z.string()).describe("States/federal"),
      stages: z.array(z.string()).describe("Introduced, hearing, floor, etc."),
    },
  },
  async ({ keywords, jurisdictions, stages }) => {
    const bills = await monitorLegislation(keywords, jurisdictions, stages);
    return { content: [{ type: "text", text: JSON.stringify(bills) }] };
  },
);
 
server.registerTool(
  "compare_versions",
  {
    description: "Compare bill versions across sessions",
    inputSchema: {
      billId: z.string(),
      versions: z.array(z.string()).describe("Version numbers to compare"),
    },
  },
  async ({ billId, versions }) => {
    const diff = await compareLegislativeVersions(billId, versions);
    return { content: [{ type: "text", text: JSON.stringify(diff) }] };
  },
);

Upload the legal brief to your assistant with instruction:
"Verify every case citation in this brief is still good law
and identify any corrections needed."

Assistant workflow:
1. Extract all citations (Midpage tool)
2. Check each citation status (Midpage citator)
3. Identify overruled/reversed cases
4. Generate citation correction report
5. Flag negative law that contradicts arguments

{
  "mcpServers": {
    "midpage": {
      "command": "npx",
      "args": ["-y", "@midpage/mcp-server"],
      "tools": [
        "search_cases",
        "verify_citation_status",
        "find_citing_cases",
        "check_negative_law"
      ],
      "env": {
        "MIDPAGE_API_KEY": "your-api-key"
      }
    },
    "google-scholar": {
      "command": "npx",
      "args": ["-y", "google-scholar-mcp"],
      "tools": ["search_scholar", "verify_citations"]
    }
  }
}

Citation verification prompt:
"For the following cases, provide Shepard's analysis:
- Smith v. Jones, 234 F.3d 567 (5th Cir. 2020)
- Brown v. Green Corp., 890 P.2d 123 (Cal. 2019)

For each case:
1. Current status (good law/overruled/limited)
2. Citing cases (positive/negative treatment)
3. Strength of authority for this proposition
4. Alternative authorities if weakened"

// citation-validator-mcp.js
import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
import { z } from "zod";
 
const server = new McpServer({ name: "citation-validator", version: "1.0.0" });
 
server.registerTool(
  "validate_citation_format",
  {
    description: "Validate and standardize citation format",
    inputSchema: {
      citation: z.string().describe("Citation to validate"),
      jurisdiction: z.string().describe("State/federal"),
      citationStyle: z.enum(["bluebook", "alwd", "local"]),
    },
  },
  async ({ citation, jurisdiction, citationStyle }) => {
    const normalized = normalizeCitation(citation, jurisdiction);
    const formatted = applyCitationStyle(normalized, citationStyle);
    const result = { original: citation, normalized, formatted, isValid: true, corrections: [] };
    return { content: [{ type: "text", text: JSON.stringify(result) }] };
  },
);
 
server.registerTool(
  "batch_validate_brief",
  {
    description: "Validate all citations in document",
    inputSchema: {
      documentText: z.string(),
      citationStyle: z.string(),
    },
  },
  async ({ documentText, citationStyle }) => {
    const citations = extractCitations(documentText);
    const validated = await Promise.all(citations.map((c) => validateCitationFormat(c, citationStyle)));
    const summary = {
      totalCitations: citations.length,
      validCitations: validated.filter((v) => v.isValid).length,
      issues: validated.filter((v) => !v.isValid),
      correctedDocument: reconstructDocumentWithCorrections(documentText, validated),
    };
    return { content: [{ type: "text", text: JSON.stringify(summary) }] };
  },
);

{
  "mcpServers": {
    "pacer": {
      "command": "npx",
      "args": ["-y", "@open-legal-tools/pacer-mcp"],
      "env": {
        "PACER_USERNAME": "your-pacer-login",
        "PACER_PASSWORD": "your-pacer-password",
        "PACER_CLIENT_CODE": "your-client-code"
      }
    },
    "recap": {
      "command": "npx",
      "args": ["-y", "@free-law-project/recap-mcp"],
      "tools": [
        "search_pacer_documents",
        "get_docket",
        "set_filing_alerts",
        "get_document"
      ]
    }
  }
}

Workflow: Competitive intelligence monitoring

Setup alert for:
- All cases involving Company X in federal courts
- Parties: Company X (plaintiff or defendant)
- Date range: Last 2 years
- Alert frequency: Real-time

Assistant multi-source search:
1. Query PACER for Company X docket entries
2. Search RECAP for public documents
3. Cross-reference state court records
4. Extract defendants/plaintiffs for analysis
5. Generate competitive threat assessment

# Search patterns for 850M+ document database
# Example: Patent litigation involving software
 
curl -X POST https://api.recap.org/search \
  -H "Content-Type: application/json" \
  -d '{
    "query": "software patent infringement",
    "courts": ["all_federal"],
    "date_filed_after": "2023-01-01",
    "document_types": ["complaint", "judgment"],
    "limit": 1000
  }'

Monitor docket: Smith v. Jones, Case No. 2023-CV-12345

Daily workflow:
1. Check for new filings (PACER alert)
2. Extract filing metadata (date, filer, document type)
3. Download documents if Rule 34 request received
4. Flag critical events (depositions, trial dates)
5. Summarize for team

Configuration:
{
  "monitored_dockets": [
    {
      "case_id": "2023-CV-12345",
      "court": "Northern District of California",
      "check_frequency": "daily",
      "alerts": {
        "critical_events": true,
        "discovery_deadlines": true,
        "expert_disclosures": true,
        "all_filings": false
      }
    }
  ]
}

// party-tracker-mcp.js
import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
import { z } from "zod";
 
const server = new McpServer({ name: "party-tracker", version: "1.0.0" });
 
server.registerTool(
  "track_party",
  {
    description: "Monitor all cases involving a party",
    inputSchema: {
      partyName: z.string().describe("Individual or company name"),
      role: z.enum(["plaintiff", "defendant", "all"]),
      courts: z.array(z.string()).describe("Courts to monitor"),
    },
  },
  async ({ partyName, role, courts }) => {
    const cases = await searchAllCourts(partyName, role, courts);
    const dockets = await Promise.all(cases.map((c) => getDocket(c.caseId)));
    const report = {
      totalCases: cases.length,
      activeCases: cases.filter((c) => c.status === "open"),
      dockets,
      competitiveAnalysis: analyzePartyHistory(cases),
    };
    return { content: [{ type: "text", text: JSON.stringify(report) }] };
  },
);

Research request:
"My client is negotiating a software licensing agreement
with a major tech company. Research the landscape of:
1. Recent software licensing disputes
2. Key contract provisions courts scrutinize
3. Best practices from firm precedents
4. Relevant patent litigation affecting licensing
5. Synthesize into deal negotiation memo"

Assistant agent workflow:
Step 1: Research Phase
  - Query Midpage for software licensing cases (last 5 years)
  - Extract critical issues from leading cases

Step 2: Precedent Phase
  - Search SharePoint for firm's prior licensing agreements
  - Extract successful clause language
  - Identify deal-breaker provisions by client type

Step 3: Patent Phase
  - Search Google Patents for relevant technology patents
  - Identify software licensing implications
  - Flag cross-license requirements

Step 4: Synthesis Phase
  - Organize findings by provision (IP ownership, liability, etc.)
  - Cross-reference case law to deal terms
  - Generate risk assessment matrix

Step 5: Deliverable Phase
  - Create negotiation memo with:
    * Market landscape summary
    * Key case takeaways
    * Recommended provisions
    * Risk mitigation strategies

{
  "mcpServers": {
    "midpage": { },
    "sharepoint": { },
    "google-patents": {
      "command": "npx",
      "args": ["-y", "google-patents-mcp"]
    },
    "courtlistener": { }
  },
  "research_workflows": {
    "licensing_negotiation": {
      "sources": ["midpage", "sharepoint", "google-patents"],
      "steps": [
        "search_cases",
        "extract_precedents",
        "patent_analysis",
        "synthesize_findings"
      ],
      "output_format": "negotiation_memo"
    }
  }
}

User: "I need to understand California's position on
non-compete agreements in employment contracts.
What should I research?"

Assistant research planner:
"I'll develop a research plan covering:

Phase 1: Controlling Authority
  - California Supreme Court cases on non-competes
  - Test: Is non-compete reasonable? (Blue Pencil doctrine)
  - Key cases: Silguero, Edwards, Morbi

Phase 2: Legislature Response
  - California Business & Professions Code § 16600
  - Legislative amendments (2024 updates)
  - Statutory interpretation

Phase 3: Modern Applications
  - Recent Court of Appeal decisions (2023-2025)
  - Non-compete vs. non-solicitation distinction
  - Remote work implications

Phase 4: Practical Synthesis
  - Enforceable provisions despite § 16600
  - Trade secret protection alternatives
  - Employment agreement drafting best practices

Estimated research time: 2-3 hours
Key sources: Midpage, Clio firm memos, Bar Association guidance"

# research_refinement_pattern.py
research_iteration = {
    "initial_query": "California non-compete enforceability",
 
    "iteration_1": {
        "findings": "Initial results show non-competes generally unenforceable",
        "questions_raised": [
            "What about legitimate business interests exception?",
            "How does this apply to C-suite executives?",
            "Recent trends in remote work?"
        ]
    },
 
    "iteration_2": {
        "refined_searches": [
            "California non-compete exceptions legitimate business interests",
            "Non-compete enforcement executive employment contracts",
            "Non-compete non-solicitation remote work 2024"
        ],
        "new_findings": "Narrow exceptions for limited legitimate interests",
        "additional_questions": [
            "What qualifies as legitimate business interest?",
            "How do courts balance employee freedom with employer protection?"
        ]
    },
 
    "iteration_3": {
        "targeted_research": [
            "California Business Code § 16600 case law interpretation",
            "Blue pencil doctrine in non-compete context",
            "Employer strategic alternatives to non-competes"
        ],
        "synthesis": "Actionable guidance for client"
    }
}

Query: "Judge Smith handles patent licensing disputes
in the Northern District of California. What patterns
can I identify in her decisions?"

Analytics workflow:
1. Search RECAP for all Judge Smith cases
2. Extract relevant patent licensing decisions
3. Analyze patterns:
   - Grant rate on summary judgment motions
   - Treatment of injunction requests
   - Damages awards (high/low/reasonable)
   - Attitude toward expert witnesses
4. Compare to district averages
5. Generate litigation strategy recommendations

{
  "mcpServers": {
    "judicial-analytics": {
      "command": "npx",
      "args": ["-y", "judicial-analytics-mcp"],
      "tools": [
        "judge_profile",
        "decision_patterns",
        "motion_grant_rates",
        "remedy_analysis",
        "compare_judges"
      ],
      "env": {
        "RECAP_API": "api-key",
        "COURTLISTENER_API": "api-key"
      }
    }
  }
}

Workflow: Opposing Counsel Analysis

For: Attorney Jane Doe (plaintiff's counsel)

Metrics extracted:
- Settlement rate: 45% (vs. district average 52%)
- Trial win rate: 62% (favorable)
- Average verdict: $1.2M (high-value cases)
- Expert witness preferences: Specific firms used repeatedly
- Motion practice: Aggressive discovery requests
- Deposition style: Confrontational

Strategic implications:
- Prepare for aggressive discovery
- Expect push to trial
- May be open to structured settlements
- Anticipated expert witnesses based on historical patterns

Query: "Predict the likely outcome of our employment
discrimination case with Judge Smith in NDCA."

Analysis:
1. Historical data: Judge Smith's discrimination cases
2. Factors: Strength of evidence, damages, settlement history
3. Prediction: 65% chance of plaintiff success in trial
4. Outcome distribution:
   - Summary judgment win (defendant): 20%
   - Trial win (defendant): 15%
   - Trial win (plaintiff): 45%
   - Settlement: 20%
5. Median award if plaintiff prevails: $850,000
6. Confidence interval: +/- $300,000

// settlement-analytics-mcp.js
import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
import { z } from "zod";
 
const server = new McpServer({ name: "settlement-analytics", version: "1.0.0" });
 
server.registerTool(
  "predict_settlement_range",
  {
    description: "Analyze historical settlements for range prediction",
    inputSchema: {
      caseType: z.string().describe("Employment, IP, contract, etc."),
      jurisdiction: z.string(),
      judge: z.string(),
      damages: z.number().describe("Claimed damages"),
    },
  },
  async ({ caseType, jurisdiction, judge, damages }) => {
    const historicalCases = await searchSimilarCases(caseType, jurisdiction, judge);
    const settlementRates = analyzeSettlements(historicalCases);
    const predictedRange = calculateRange(damages, settlementRates, judge);
    const output = {
      predictedLowRange: predictedRange.low,
      predictedMidRange: predictedRange.mid,
      predictedHighRange: predictedRange.high,
      confidence: predictedRange.confidence,
      comparableCases: historicalCases.slice(0, 5),
    };
    return { content: [{ type: "text", text: JSON.stringify(output) }] };
  },
);

Search for [topic] cases in [jurisdiction]
Verify citation [citation] is still good law
Find cases that cite [case name]
Check if [case] has been overruled or limited

Search CourtListener for [query]
Get full text of [case citation]
Find oral arguments in [case]
Search by judge: [judge name] in [court]

Search Congress for bills matching [criteria]
Track amendments to [statute] by year
Compare versions of [bill] across sessions
Search Federal Register for [regulation changes]
Monitor bills in [state legislature] related to [topic]

Search PACER for cases involving [party name]
Get docket for [case number] in [court]
Set alert for filings in [case number]
Search RECAP for [party] across all federal courts
Find all cases by Judge [judge name]

Verify citation [citation] is still good law
Check Shepard's for [case]: positive/negative treatment
Find alternative authorities for [proposition]
Validate citation format for [citation] (Bluebook/ALWD)

Search [system] for documents matching [criteria]
Retrieve document [name/ID] from [location]
List recent documents in [workspace/folder]

Analyze Judge [judge name] decision patterns for [case type]
Predict settlement range for [case type] before Judge [judge]
Compare outcome patterns for opposing counsel [attorney name]
Calculate motion grant rates for [judge] on [motion type]