Generated from cyanheads/mcp-ts-template
An MCP (Model Context Protocol) server for managing budgets via the PocketSmith API.
Built on the cyanheads/mcp-ts-template, this server follows a modular architecture with robust error handling, logging, and security features while providing comprehensive PocketSmith integration.
See POCKETSMITH_README.md for complete PocketSmith-specific documentation.
- This server handles highly sensitive financial data
- Your PocketSmith API key provides complete access to your financial accounts
- Only use with trusted MCP clients and secure configurations
- Enable confirmation prompts for all financial operations
- Read the complete SECURITY.md documentation
- Store API key in secure environment variables only
- Use stdio transport for single-user scenarios
- Never add financial operations to auto-approve lists
- Set up regular API key rotation (recommended: every 90 days)
- Monitor your PocketSmith account for unauthorized changes
π¨ Never share your API key or use this server on untrusted systems.
-
Install dependencies:
npm install
-
Set up environment: Copy
.env.exampleto.envand add your PocketSmith API key -
Build:
npm run build
-
Run:
npm run start:stdio # For MCP clients npm run start:http # For web applications
get_accounts- View account balances and informationget_user_summary- Comprehensive financial overview
get_transactions- Search and filter transactions with advanced criteriaget_transaction- Get detailed information about a specific transactioncreate_transaction- Add new transactions with categories and notesupdate_transaction- Modify existing transactions (change payee, amount, category, etc.)delete_transaction- Remove transactions (with confirmation)
get_categories- View spending categories and hierarchiescreate_category- Add new spending categories with colors and parent relationshipsupdate_category- Modify existing categories (rename, recolor, reparent)
get_budgets- Analyze budget vs actual spending performanceget_budget_summary- Detailed budget analysis with period breakdowns
get_recurring_events- View scheduled/recurring transactions and income
get_transaction_attachments- View attachments for specific transactionsget_user_attachments- View all user attachments across transactionscreate_transaction_attachment- Add receipts and documents to transactions
get_category_rules- View automatic categorization rulescreate_category_rule- Create rules for auto-categorizing transactions
get_account_transactions- Get transactions for specific accountsget_transaction_accounts- Detailed transaction account information
get_currencies- View supported currencies and their formatsget_institutions- View connected financial institutions
For Claude Desktop or other MCP clients:
{
"mcpServers": {
"pocketsmith": {
"command": "node",
"args": ["/path/to/pocketsmith-mcp/dist/index.js"],
"env": {
"POCKETSMITH_API_KEY": "your_api_key_here"
}
}
}
}Built with β€οΈ and the Model Context Protocol
The following documentation is from the original cyanheads/mcp-ts-template:
Jumpstart your Model Context Protocol (MCP) development with this comprehensive TypeScript Template for building autonomous agents, servers, and clients.
This template provides a solid, beginner-friendly foundation for building all components of the MCP ecosystem, adhering to the MCP 2025-06-18 specification. It includes a powerful agent framework, a fully-featured server, a robust client, production-ready utilities, and clear documentation to get you up and running quickly.
This template is organized into three primary, interconnected components:
- π€ Agent (
src/agent/): An autonomous agent framework. The agent can connect to multiple MCP servers, discover their tools, and use them to accomplish complex tasks based on a user's prompt. Use this as a starting point for your agents. - π MCP Server (
src/mcp-server/): An extensible MCP server that can host custom tools and resources, making them available to agents and other clients. - π» MCP Client (
src/mcp-client/): A robust client for connecting to and interacting with any MCP-compliant server. The agent uses this client to communicate with the outside world.
| Feature Area | Description | Key Components / Location |
|---|---|---|
| π€ Agent Framework | Core Agent class and CLI for running autonomous agents that connect to MCP servers and use their tools to achieve goals. |
src/agent/ |
| π MCP Server | Functional server with example tools (EchoTool, CatFactFetcher) and an EchoResource. Supports stdio and Streamable HTTP transports. |
src/mcp-server/ |
| π» MCP Client | Working client aligned with MCP 2025-03-26 spec. Connects via mcp-config.json. Includes detailed comments and isolated connection management. |
src/mcp-client/ |
| π Production Utilities | Logging, Error Handling, ID Generation, Rate Limiting, Request Context tracking, Input Sanitization. | src/utils/ |
| π Type Safety/Security | Strong type checking via TypeScript & Zod validation. Built-in security utilities (sanitization, auth middleware for HTTP). | Throughout, src/utils/security/, src/mcp-server/transports/auth/ |
| βοΈ Error Handling | Consistent error categorization (BaseErrorCode), detailed logging, centralized handling (ErrorHandler). |
src/utils/internal/errorHandler.ts, src/types-global/ |
| π Documentation | Comprehensive README.md, structured JSDoc comments, API references. |
README.md, Codebase, tsdoc.json, docs/api-references/ |
| π΅οΈ Interaction Logging | Captures raw requests and responses for all external LLM provider interactions to a dedicated interactions.log file for full traceability. |
src/utils/internal/logger.ts |
| π€ Agent Ready | Includes a .clinerules developer cheatsheet tailored for LLM coding agents. | .clinerules |
| π οΈ Utility Scripts | Scripts for cleaning builds, setting executable permissions, generating directory trees, and fetching OpenAPI specs. | scripts/ |
| Services | Reusable modules for LLM (OpenRouter) and data storage (DuckDB) integration, with examples. | src/services/, src/storage/duckdbExample.ts |
This template is already powering several MCP servers, demonstrating its flexibility and robustness:
| Project | Description |
|---|---|
| clinicaltrialsgov-mcp-server | Provides an LLM-friendly interface to the official ClinicalTrials.gov v2 API, enabling agents to analyze clinical study data. |
| pubmed-mcp-server | Enables AI agents to search, retrieve, and visualize biomedical literature from PubMed via NCBI E-utilities. |
| git-mcp-server | Provides an enterprise-ready MCP interface for Git operations, allowing agents to manage repositories programmatically. |
| obsidian-mcp-server | Allows AI agents to read, write, search, and manage notes in Obsidian via the Local REST API plugin. |
| atlas-mcp-server | An advanced task and knowledge management system with a Neo4j backend for structured data organization. |
| filesystem-mcp-server | Offers platform-agnostic file system capabilities for AI agents, including advanced search and directory traversal. |
| workflows-mcp-server | A declarative workflow engine that allows agents to execute complex, multi-step automations from simple YAML files. |
Note: toolkit-mcp-server was built on an older version of this template and is pending updates.
You can also see my GitHub profile for additional MCP servers I've created.
Clone the repository and install dependencies:
git clone https://github.com/cyanheads/mcp-ts-template.git
cd mcp-ts-template
npm installnpm run build
# Or use 'npm run rebuild' for a clean installYou can run the included MCP server to make its tools available.
- Via Stdio (Default):
npm run start:server
- Via Streamable HTTP:
npm run start:server:http
The agent can be run from the command line to perform tasks. It will automatically connect to the servers defined in src/mcp-client/client-config/mcp-config.json. If running the agent, you must have the MCP config set up correctly and your openrouter API key configured in .env.
npm run start:agent "Your prompt here"
# Example:
npm run start:agent "Use the echo tool to say hello world and then get a cat fact."Configure the MCP server's behavior using these environment variables:
| Variable | Description | Default |
|---|---|---|
MCP_TRANSPORT_TYPE |
Server transport: stdio or http. |
stdio |
MCP_HTTP_PORT |
Port for the HTTP server (if MCP_TRANSPORT_TYPE=http). |
3010 |
MCP_HTTP_HOST |
Host address for the HTTP server (if MCP_TRANSPORT_TYPE=http). |
127.0.0.1 |
MCP_ALLOWED_ORIGINS |
Comma-separated allowed origins for CORS (if MCP_TRANSPORT_TYPE=http). |
(none) |
MCP_AUTH_MODE |
Authentication mode for HTTP: jwt (default) or oauth. |
jwt |
MCP_AUTH_SECRET_KEY |
Required for jwt mode. Secret key (min 32 chars) for signing/verifying auth tokens. |
(none - MUST be set in production) |
OAUTH_ISSUER_URL |
Required for oauth mode. The issuer URL of your authorization server. |
(none) |
OAUTH_AUDIENCE |
Required for oauth mode. The audience identifier for this MCP server. |
(none) |
OPENROUTER_API_KEY |
API key for OpenRouter.ai service. Required for the agent to function. | (none) |
The agent uses the MCP client to connect to servers. This is configured in src/mcp-client/client-config/mcp-config.json. You must list all MCP servers the agent should connect to in this file.
For a detailed guide, see the Client Configuration README.
src/agent/: Contains the core agent framework, including theAgentclass and a CLI for running the agent.src/mcp-client/: Implements the MCP client logic for connecting to and interacting with external MCP servers.src/mcp-server/: Contains the MCP server implementation, including example tools, resources, and transport handlers.src/config/: Handles loading and validation of environment variables and application configuration.src/services/: Provides reusable modules for integrating with external services (DuckDB, OpenRouter).src/types-global/: Defines shared TypeScript interfaces and type definitions.src/utils/: A collection of core utilities (logging, error handling, security, etc.).src/index.ts: The main entry point for the application, responsible for initializing and starting the MCP server.
Explore the full structure yourself:
See the current file tree in docs/tree.md or generate it dynamically:
npm run treeFor detailed guidance on how to add your own custom Tools and Resources to the MCP server, please see the Server Extension Guide.
The agent's core logic is in src/agent/agent-core/agent.ts. You can modify its system prompt, the models it uses (google/gemini-2.5-flash by default), and its decision-making loop to change its behavior.
Looking for more examples, guides, and pre-built MCP servers? Check out the companion repository:
β‘οΈ cyanheads/model-context-protocol-resources
This project is licensed under the Apache License 2.0. See the LICENSE file for details.