MCP 服务器

filevine-mcp

An MCP server that exposes Filevine matter management and Lead Docket intake APIs as natural-language-callable tools, enabling legal teams to query, update, and create projects and leads directly from Claude.

README

filevine-mcp

An open-source MCP (Model Context Protocol) server for Filevine matter management and Lead Docket intake.

Built and maintained by Advocate Cloud Solutions — a Salesforce consultancy for law firms. This MCP exposes the Filevine + Lead Docket API surface as Claude-callable tools so legal teams can query, build, and fix matters and intake leads through natural language — running entirely on your own machine against your own Filevine org.

What this is

Filevine is the leading mid-market personal injury / mass tort case management platform (~3,000 firms, $400M Series funding Sept 2025). Lead Docket is Filevine's acquired intake/CRM product — leads from web forms, chatbots, and calls, with scoring, routing, and conversion into Filevine projects.

filevine-mcp is a standalone, single-tenant MCP server. You generate your own Filevine PAT, run the server locally (or on infrastructure you control), and connect it to any MCP-compatible client — Claude Desktop, Claude Code, Cowork, or anything else that speaks MCP.

This MCP wraps both Filevine and Lead Docket under one namespace so a single install gives Claude access to the full intake-to-matter funnel.

Three workflows the MCP unlocks:

Pull — "Show me every Lead Docket lead from this week's PPC campaigns that hasn't been contacted yet, plus every Filevine project in pre-litigation phase older than 90 days."
Fix — "On project 78213, the policy-limit field is wrong — should be $250K not $25K. Update it and add a note. Also re-route the bad-faith Lead Docket lead to the senior PI partner."
Build — "Convert this Lead Docket lead into a new Filevine PI project, pre-fill the intake section from the lead's form data, and schedule the consultation."

Quick start

git clone https://github.com/Advocate-Cloud-Solutions/filevine-mcp.git
cd filevine-mcp
npm install
cp .env.example .env  # fill in your Filevine PAT, client_id, client_secret
npm run build

Then add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "filevine": {
      "command": "node",
      "args": ["/absolute/path/to/filevine-mcp/dist/index.js"],
      "env": {
        "FILEVINE_CLIENT_ID": "...",
        "FILEVINE_CLIENT_SECRET": "...",
        "FILEVINE_PAT": "...",
        "FILEVINE_USER_ID": "...",
        "FILEVINE_ORG_ID": "..."
      }
    }
  }
}

See AGENTS.md for the full setup walkthrough including how to generate a PAT in the Filevine Org Admin tools.

Architecture at a glance

PAT exchange auth for Filevine; separate API key for Lead Docket (both read from environment variables)
Dynamic schema introspection — Filevine's per-firm Project Type templating means every write tool needs runtime schema discovery (1hr in-memory cache, force_refresh flag)
Rate-limit governor — Per-endpoint token buckets (Filevine: 320/min general, 250/min billing, 5/min Reports + Vitals; Lead Docket: separate buckets per its own limits)
Optional audit logging — JSONL audit log per ABA Opinion 512, off by default, opt-in via AUDIT_LOG_PATH

Full architecture: see ARCHITECTURE.md. Lead Docket-specific design notes: docs/LEAD_DOCKET.md.

Tool surface

~32 tools total: 25 Filevine core + 7 Lead Docket.

Filevine core (25 tools)

Tool	What it does
`filevine_search_projects`	Query projects by phase, project type, assignee, date
`filevine_describe_sections`	Returns full schema (sections + fields + types) for a project type
`filevine_get_section` / `filevine_update_section`	Read/write static section field values
`filevine_create_collection_item`	Add a row to a Collection section (medical providers, defendants, etc.)
`filevine_create_note`	Add a note with @mentions and attachments
`filevine_run_report`	Execute a saved Periscope report (heavily cached due to 5/min cap)

Lead Docket (7 tools)

Tool	What it does
`filevine_lead_search`	Query leads by status, source, owner, date range
`filevine_lead_get`	Get a lead's full intake-form data + scoring
`filevine_lead_create`	Create a lead (manual entry; web forms write via Lead Docket UI)
`filevine_lead_update_status`	Move a lead through statuses (new → contacted → qualified → converted/declined)
`filevine_lead_assign`	Reassign a lead to a different intake specialist or attorney
`filevine_lead_convert_to_project`	Convert a qualified lead into a Filevine project; carries form data into intake section
`filevine_lead_list_sources`	List lead sources (PPC campaigns, referrals, web forms) for routing decisions

Full inventory: docs/TOOLS.md.

Roadmap

6-phase build, ~17-25 dev-weeks for v1.

Recon + sandbox (1-2 wk) — engineering seats, validate endpoints, draft type system
Filevine read-only MCP (4-6 wk) — 10 read tools, PAT exchange, rate-limit governor
Filevine write tools (3-4 wk) — Section writes, note/task create, collection CRUD, doc upload
Filevine reports + caching (4-6 wk) — Reports tool, materialized cache, Periscope passthrough
Lead Docket sub-module (2-3 wk) — 7 tools, separate auth, lead-to-project conversion bridge
Hardening + CPI (3-4 wk) — Certified Partner Integration application, security review, error taxonomy, observability

Full roadmap: ROADMAP.md.