Legal Docket Monitor MCP Server
Wraps the CourtListener API for real-time federal docket monitoring, with pluggable client-list connectors (SharePoint, Excel, webhook) and tools for docket monitoring and client matching via elicitation.
README
Legal Docket Monitor — MCP Server
An MCP server that wraps the CourtListener API for real-time federal docket monitoring — and, when configured, automatically connects new filings to your firm's existing client list, wherever it lives.
Built for litigation teams who don't want to manually check PACER for new filings, and who want that monitoring connected to "which client does this affect" without re-keying data between systems.
Why this exists
Most docket-monitoring tools are single-purpose: they watch a case and email you. This is an MCP server instead — a tool layer any AI agent (Claude Desktop, Cursor, your own orchestrator) can call. That means an agent can watch dockets, triage what's urgent, and link filings to clients as part of a larger workflow you control, not a fixed pipeline someone else built.
What's new in v2
v1 monitored dockets. v2 adds:
- A pluggable client-list connector — SharePoint List, Excel Table, or your own backend via webhook, all behind one interface. Pick whichever fits your stack; switching later is a config change, not a rewrite.
- Sampling-based filing triage — new filings are automatically classified (routine / needs-review / urgent) using the connected AI client's own model. The server never holds a model API key.
- Elicitation-based client matching — when a filing's case name could match more than one client, the server asks a human to confirm instead of guessing.
Tools
Docket monitoring
| Tool | What it does |
|---|---|
search_dockets |
Search for cases by name, party, or keyword |
get_docket |
Full metadata for a docket (judge, dates, court, status) |
get_new_filings |
Entries since a given date |
get_docket_summary |
Metadata + last N filings in one call |
get_parties |
Parties and attorneys for a docket |
watch_docket |
Add a docket to the watch list |
unwatch_docket |
Remove from the watch list |
list_watched_dockets |
See all tracked matters |
check_watched_dockets |
Poll all watched dockets; triages new filings by urgency via sampling |
Client matching (requires a configured client source — see below)
| Tool | What it does |
|---|---|
find_matching_clients |
Fuzzy-search the connected client list by name |
link_docket_to_client |
Link a docket to a client — auto-matches if unambiguous, asks via elicitation if not |
check_client_source |
Health check the configured backend |
Architecture
MCP tool layer (src/index.ts)
│
├─→ CourtListener client (src/courtlistener.ts) — docket data
├─→ Watch list (src/watchlist.ts) — tracked matters
├─→ Triage (src/triage.ts) — sampling-based urgency classification
├─→ Client matching (src/client-matching.ts) — elicitation-based resolution
└─→ ClientListSource (src/client-sources/) — pluggable client-list backend
├─ SharePointListSource (Microsoft Graph: SharePoint Lists)
├─ ExcelTableSource (Microsoft Graph: Excel Workbook API)
└─ WebhookSource (your own REST endpoint)
Every client-list backend implements the same ClientListSource interface. The tool layer never knows which one is active — that's controlled entirely by environment configuration (see below). This is what makes "connect your client list however you want" an actual property of the system rather than a roadmap item.
Prerequisites
- Node.js 18+
- A free CourtListener account, for an API token: https://www.courtlistener.com/sign-in/
- If connecting SharePoint or Excel: an Entra ID (Azure AD) app registration with
Sites.Read.Write.All(or narrower, scoped to the specific site) Graph API permissions, using the client-credentials flow - If connecting a custom backend: an HTTP endpoint matching the contract below
Installation
git clone <your-repo>
cd legal-docket-monitor
npm install
npm run build
Configuration
Docket monitoring works standalone with just a CourtListener token. Client matching requires one client-source configuration block, chosen by CLIENT_SOURCE_TYPE.
# Required for docket monitoring
COURTLISTENER_API_TOKEN=your_token_here
# Optional — enables client-matching tools. One of: sharepoint, excel, webhook
CLIENT_SOURCE_TYPE=
# --- SharePoint List ---
SHAREPOINT_SITE_ID=
SHAREPOINT_LIST_ID=
GRAPH_TENANT_ID=
GRAPH_CLIENT_ID=
GRAPH_CLIENT_SECRET=
# --- Excel Table (same Graph credentials as SharePoint) ---
EXCEL_DRIVE_ID=
EXCEL_ITEM_ID=
EXCEL_TABLE_NAME=
GRAPH_TENANT_ID=
GRAPH_CLIENT_ID=
GRAPH_CLIENT_SECRET=
# --- Custom webhook (your own code) ---
WEBHOOK_BASE_URL=
WEBHOOK_AUTH_TOKEN=
WEBHOOK_SEARCH_FALLBACK=true # set false to require the backend implement /clients/search
If CLIENT_SOURCE_TYPE is unset, the server runs in docket-monitoring-only mode — find_matching_clients and link_docket_to_client report that no client source is configured, rather than failing.
SharePoint List — expected columns
| Column | Maps to | Required |
|---|---|---|
Title |
clientName |
Yes |
Aliases |
comma-separated alternate names | No |
MatterIds |
comma-separated docket numbers (also where links get written) | No |
ContactEmail |
— | No |
PracticeArea |
— | No |
Excel Table — expected columns
Same shape as above, but the data must be formatted as an actual Excel Table (Insert → Table), not a raw range — the Graph Workbook API operates on named tables. Header names: ClientName, Aliases, MatterIds, ContactEmail, PracticeArea.
Custom webhook — expected contract
Your backend implements:
GET /clients → ClientRecord[]
GET /clients/search?q=&limit= → ClientMatch[] (optional; falls back to local
fuzzy search over GET /clients if omitted)
POST /clients/:id/link → body: { docketId: number }, 204 on success
GET /health → { ok: boolean, detail?: string }
All requests carry Authorization: Bearer <WEBHOOK_AUTH_TOKEN>. See src/client-sources/types.ts for the exact ClientRecord and ClientMatch shapes.
Connecting to Claude Desktop
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"legal-docket-monitor": {
"command": "node",
"args": ["/absolute/path/to/legal-docket-monitor/dist/index.js"],
"env": {
"COURTLISTENER_API_TOKEN": "your_token_here",
"CLIENT_SOURCE_TYPE": "sharepoint",
"SHAREPOINT_SITE_ID": "...",
"SHAREPOINT_LIST_ID": "...",
"GRAPH_TENANT_ID": "...",
"GRAPH_CLIENT_ID": "...",
"GRAPH_CLIENT_SECRET": "..."
}
}
}
}
Restart Claude Desktop after editing. Note: sampling and elicitation require a client that supports those MCP capabilities — check your client's documentation if check_watched_dockets triage or link_docket_to_client confirmation prompts don't appear.
Example agent interactions
Watch a matter and check in on it:
→ watch_docket(docket_id=4214664, matter_id="smith-v-jones")
→ check_watched_dockets()
→ new filings are automatically triaged: 🔴 urgent / 🟡 needs-review / ⚪ routine
Connect a docket to your client list:
→ link_docket_to_client(docket_id=4214664, case_name="Acme Corp Holdings LLC v. Smith")
→ if one clear match: linked automatically
→ if ambiguous: you're asked to pick which client record is correct
Check whether your client source is set up correctly:
→ check_client_source()
Running tests
npm test # full suite — 55 tests across docket, watchlist, and client-source logic
npm run typecheck # type-check only, faster than a full build
All tests mock network calls (fetch) — no real CourtListener, Graph, or webhook requests are made during testing.
Project structure
src/
index.ts MCP server entry point — all tool definitions
courtlistener.ts CourtListener API v4 client
watchlist.ts In-memory watch list
triage.ts Sampling-based filing urgency triage
client-matching.ts Elicitation-based docket→client resolution
client-sources/
types.ts ClientListSource interface + fuzzy matching + caching
graph-client.ts Shared Microsoft Graph auth client
sharepoint-source.ts SharePoint List adapter
excel-source.ts Excel Table adapter
webhook-source.ts Custom backend adapter
factory.ts Picks the adapter from CLIENT_SOURCE_TYPE
*.test.ts Vitest suites
CLAUDE.md Guidance for AI agents extending this codebase
Limitations & production considerations
- Watch list is in-memory — lost on restart. Swap
WatchListfor a persistent store for production use. - CourtListener coverage isn't complete for every federal case — cases not submitted to RECAP may have limited entry data.
- Client-source caching: client lists are cached for 5 minutes by default (
TtlCacheintypes.ts) to avoid hammering SharePoint/Excel/your webhook on every fuzzy search. Adjust if your client list changes more frequently than that. - Sampling and elicitation require client support. Not every MCP client implements these capabilities yet. The server degrades gracefully (triage falls back to raw filing text; matching without elicitation support will simply not offer disambiguation), but won't error out.
- No webhooks from CourtListener (yet) —
check_watched_docketsis poll-based. A future extension could register CourtListener webhooks for push-based alerts instead.
Contributing
See CLAUDE.md for architecture conventions, the adapter-extension checklist, and PR expectations if you're adding a new client-list backend or tool.
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.