Developers · MCP

Jupid MCP

Connect Claude, Cursor, and other MCP clients to your Jupid workspace. Read-only access to transactions, counterparties, and Schedule C cash-flow reports — returned as structured JSON to whatever AI agent you’re using.

Remote endpoint

Looking for the non-technical tour? See the MCP overview.

Overview

The Jupid MCP server exposes six read-only tools over the Model Context Protocol. An MCP-capable client — Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, ChatGPT, or any custom agent — can call these tools on the user’s behalf to answer questions grounded in live Jupid data.

What you can do: view a user’s Jupid profile ( me), search and inspect transactions ( search_transactions, get_transaction), look up vendors and clients ( search_counterparties, get_counterparty), and generate Schedule C cash-flow reports ( get_cashflow_report).

Requirements: an active Jupid account (sign up at app.jupid.com) and an MCP client. Authentication happens through OAuth 2.0 on first use — no secrets to paste.

Transport: streamable HTTP, stateless per request. Clients that only speak stdio can use mcp-remote as a proxy.

Quickstart

Pick your client. Each tab shows the exact config and where it lives. OAuth runs the first time a tool is called — most clients open a browser window automatically.

Add Jupid to Claude Desktop by editing claude_desktop_config.json. Uses mcp-remote as a lightweight proxy — Claude Desktop does not yet speak streamable HTTP natively.

Open claude_desktop_config.json. On macOS it lives at ~/Library/Application Support/Claude/claude_desktop_config.json. On Windows: %APPDATA%\Claude\claude_desktop_config.json.

claude_desktop_config.jsonjson
{
  "mcpServers": {
    "jupid": {
      "command": "npx",
      "args": ["mcp-remote", "https://backend.jupid.com/mcp"]
    }
  }
}

Restart Claude Desktop. On first tool use, Claude will prompt you to complete OAuth in the browser. If you prefer a static token (see Authentication below), add an AUTH_HEADER env var with your Bearer token.

Authentication

Jupid MCP supports two authentication paths. OAuth 2.0 is the default and works for every client listed above. Bearer tokens are a fallback for headless agents, CI jobs, and custom integrations that cannot open a browser for consent.

OAuth 2.0 (recommended)

Authorization code flow with PKCE (S256) plus Dynamic Client Registration (RFC 7591). Any compliant MCP client can connect without pre-registration — Jupid issues a client ID on the fly.

Discovery follows the MCP spec: WWW-Authenticate: Bearer resource_metadata="…" on the first 401, then the client fetches the .well-known/oauth-protected-resource document, which points to the authorization server. No manual configuration on your side.

Bearer token (headless)

Generate a token in app.jupid.com → Integrations → AI Agents. Tokens are prefixed jp_ and can be revoked at any time.

Pass the token in the Authorization header:

bash
curl https://backend.jupid.com/mcp \
    -H "Authorization: Bearer jp_..." \
    -H "Content-Type: application/json" \
    -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Tokens are scoped to a single Jupid account and inherit that user’s read-only permissions. Revoking the token immediately invalidates any agent using it.

Available tools

All six tools are registered with readOnlyHint: true and return structured JSON. Amounts are numbers in US dollars (negative for expenses), dates are ISO 8601, and categories use Schedule C line references where applicable.

User

Transactions

Counterparties

Reports

Example prompts

Representative prompts that exercise the six tools. Share these with users as starting points — the agent composes the tool call itself.

Freelancer

Fast answers to tax and budget questions without leaving the editor.

  • What did I spend on software this quarter?
  • Show me all Stripe charges over $500 from the last 90 days.
  • How much did I pay in contract labor in Q1?
  • Which subscriptions am I paying for that I haven’t used in 60 days?

Small-business owner

Cross-cutting views across clients, vendors, and categories.

  • Generate a Schedule C cash flow report for Q1 by category.
  • List my top 10 counterparties by spend in the last 12 months.
  • Summarize all transactions with Mercury this year.
  • What’s my net margin month-over-month since January?

Accountant

Verify a client’s books without screen-sharing a dashboard.

  • Flag any transaction over $2,000 in the last 30 days that doesn’t have a counterparty.
  • Break down Q1 expenses by Schedule C line and show the top three vendors per line.
  • Find all income transactions with no matching invoice narrative.
  • List pending transactions older than 7 days.

Security & data handling

  • Read-only by design. Every tool is registered with readOnlyHint: true. The server exposes no mutating operations — an agent cannot create transactions, edit counterparties, or change account settings through MCP.
  • User-scoped queries. Every query filters by user_id and is enforced at the database layer via row-level security. Tokens cannot cross accounts.
  • Filtered response fields. JSONB metadata, context, and narrative fields are filtered to safe public subsets. Internal scoring, reasoning notes, confidence metrics, and enrichment sources are never returned over MCP.
  • No AI-provider training. Data flows from Jupid directly to the user’s MCP client. Jupid does not share MCP-accessed data with AI providers for training or improvement.
  • Rate limit. 100 tool calls per hour per token on standard plans. The server returns 429 with a Retry-After header. Contact support@jupid.com for higher limits.
  • Privacy policy. See the Jupid privacy policy for data collection, retention, and deletion terms, including MCP-specific handling.

Troubleshooting

Resources

Feedback, bugs, or integration questions? Email support@jupid.com.