helcim-mcp
Read-only MCP server for the Helcim payment API, letting AI agents safely query customers, invoices, card transactions, subscriptions, and more without risking financial mutations.
README
helcim-mcp
Unofficial community MCP server and developer toolkit for the Helcim API. Secure, typed, agent-friendly, and read-only by default. Not affiliated with, sponsored by, maintained by, or endorsed by Helcim Inc.
helcim-mcp is a production-quality, TypeScript monorepo that makes it safe
and easy for AI agents (and humans) to work with the Helcim payment platform.
It ships three things:
@helcim-mcp/server- an MCP server exposing read-only Helcim tools (customers, invoices, card transactions, card batches, recurring payment plans, subscriptions, connection test).@helcim-mcp/core- a typed, idempotency-aware Helcim API client with normalized errors, rate-limit handling, and secret redaction.@helcim-mcp/webhooks- a standalone Helcim webhook verifier (HMAC-SHA256 signature verification, timestamp validation, replay protection, typed events).
Why would you use this?
- You want an AI agent to answer questions about your Helcim data - "what invoices are outstanding?", "show recent card transactions", "find the customer for this invoice", "which subscriptions need attention?" - without ever risking a financial mutation.
- You want a clean, typed Helcim client that handles the API's quirks
(HTTP 200 ≠ success,
errorsobject shapes, idempotency, rate limits, pagination) so you don't have to. - You want to verify Helcim webhooks securely with constant-time signature comparison and replay protection, without re-inventing the HMAC scheme.
The MCP server is read-only by default. It physically cannot create, update, delete, or move money - there are no such tools. Even a token with full processing privileges cannot trigger a financial mutation through this server.
Quick start
1. Get a Helcim API token
Log in to your Helcim account (or a developer test account), go to All Tools → Integrations → API Access Configurations, and create a configuration. For read-only use, set General: Read, Settings: Read, and Transaction Processing: None.
2. Run the MCP server
# From source
git clone https://github.com/tejasghalsasi/helcim-mcp.git
cd helcim-mcp
pnpm install
pnpm rebuild esbuild # required: pnpm 11 blocks esbuild's postinstall by default
pnpm build
# Set your token (never commit it)
export HELCIM_API_TOKEN="your_token_here"
# Run over stdio
node packages/mcp/dist/index.js
2b. Run with Docker (optional)
docker build -t helcim-mcp .
docker run --rm -e HELCIM_API_TOKEN=your_token_here helcim-mcp
2c. Run via npx (once published)
npx @helcim-mcp/server
3. Connect it to an MCP client
Add this to your MCP client config (e.g. Claude Desktop, Cursor, or any MCP client):
{
"mcpServers": {
"helcim": {
"command": "node",
"args": ["/absolute/path/to/helcim-mcp/packages/mcp/dist/index.js"],
"env": {
"HELCIM_API_TOKEN": "your_token_here"
}
}
}
}
4. Ask your agent
Once connected, your agent can call tools like:
connection_test- confirm the token works.list_invoiceswithstatus: "DUE"- "what invoices are outstanding?"list_card_transactions- "show recent card transactions."get_customer- "find the customer for this invoice."list_subscriptionswithhasFailedPayments: true- "which subscriptions need attention?"
How read-only mode works
- The MCP server exposes only read tools. There are no payment, refund, capture, reversal, withdraw, settle, or delete tools.
- The core client exposes no write methods in v1.
- If a future version adds writes, it will require an explicit
HELCIM_ENABLE_WRITES=trueenvironment variable and a separate high-risk feature flag for financial mutations, with strong documentation and tests. - HTTP 200 is not treated as success. Helcim explicitly warns that a 200
response does not mean the requested action succeeded; the client surfaces
errorsin the body as typed errors.
How credentials are protected
- The API token is read only from the
HELCIM_API_TOKENenvironment variable. Never hardcoded, never committed, never logged. - All log lines and error messages pass through
redact(). Token-like strings, card numbers, and F6L4 values are replaced with<redacted-...>. - The token is never exposed to the model. The MCP server returns only redacted data and typed error codes.
- See SECURITY.md for the full security model.
Architecture
flowchart LR
subgraph Client["MCP Client (LLM)"]
A[Agent]
end
subgraph Server["@helcim-mcp/server"]
M[MCP Server<br/>stdio transport]
T[Read-only tools<br/>13 tools]
end
subgraph Core["@helcim-mcp/core"]
C[HelcimClient]
H[HelcimHttpClient<br/>auth, idempotency,<br/>rate-limit, redaction]
E[Normalized errors]
end
subgraph Webhooks["@helcim-mcp/webhooks"]
W[HelcimWebhookVerifier<br/>HMAC-SHA256, replay protection]
end
subgraph Helcim["Helcim API"]
API[api.helcim.com/v2]
end
A -->|JSON-RPC over stdio| M
M --> T
T --> C
C --> H
H -->|HTTPS + api-token| API
W -.->|verifies signed events| API
The monorepo layout:
helcim-mcp/
├── packages/
│ ├── core/ # Typed Helcim API client (read-safe)
│ ├── mcp/ # MCP server (read-only tools)
│ ├── webhooks/ # Webhook verifier
│ └── fixtures/ # Deterministic mock responses + test vectors
├── examples/ # Copy-paste usage examples
├── docs/ # Architecture, env reference, troubleshooting
└── scripts/ # Smoke test, CI helpers
Example interaction
Agent: "What invoices are currently outstanding?"
list_invoices(status: "DUE")
→ { count: 2, invoices: [
{ invoiceId: 28658838, invoiceNumber: "INV1000", status: "DUE", currency: "CAD", customerId: 2488717 },
{ invoiceId: 28658839, invoiceNumber: "INV1001", status: "DUE", currency: "USD", customerId: 2488718 }
] }
Agent: "Show recent card transactions."
list_card_transactions(limit: 5)
→ { count: 2, transactions: [
{ transactionId: 25557533, status: "APPROVED", type: "purchase", amount: 100.99, currency: "CAD", cardType: "MC", customerCode: "CST1000" },
{ transactionId: 25557534, status: "DECLINED", type: "purchase", amount: 250.00, currency: "CAD", cardType: "VI", customerCode: "CST1001" }
] }
Agent: "Find the customer associated with this invoice."
get_invoice(invoiceId: 28658838) → { customerId: 2488717, ... }
get_customer(customerId: 2488717) → { customerCode: "CST1000", businessName: "Acme Widgets Ltd", ... }
Agent: "Show subscriptions requiring attention."
list_subscriptions(hasFailedPayments: true)
→ { count: 1, subscriptions: [ { id: 42, status: "ACTIVE", hasFailedPayments: true, customerCode: "CST1000", ... } ] }
Agent: "Process a refund for transaction 25557533."
→ Error: Unknown tool: process_refund
The agent cannot move money. There is no such tool.
Webhook verification
import { HelcimWebhookVerifier } from '@helcim-mcp/webhooks';
const verifier = new HelcimWebhookVerifier(process.env.HELCIM_VERIFIER_TOKEN!);
// In your webhook handler (e.g. Next.js route handler):
export async function POST(req: Request) {
const body = await req.text();
const headers = Object.fromEntries(req.headers.entries());
try {
const verified = verifier.verify(headers, body);
// verified.event.type === 'cardTransaction' | 'terminalCancel'
return new Response('ok', { status: 200 });
} catch (err) {
return new Response('invalid signature', { status: 401 });
}
}
See examples/webhook-nextjs.md for a full Next.js example.
Environment variables
| Variable | Required | Description |
|---|---|---|
HELCIM_API_TOKEN |
Yes (for server) | Your Helcim API token. |
HELCIM_BASE_URL |
No | Override base URL (default https://api.helcim.com/v2). |
HELCIM_DEBUG |
No | true to enable redacted request logging. |
HELCIM_TIMEOUT_MS |
No | Request timeout in ms (default 15000). |
HELCIM_VERIFIER_TOKEN |
For webhooks | Your Helcim webhook verifier token. |
See docs/environment.md for the full reference.
Reference
- Supported tools & API matrix - every MCP tool and core client method mapped to its Helcim endpoint.
- Architecture - design decisions and data flow.
- Troubleshooting - common issues and fixes.
- Environment reference - all env vars.
- Release process - how to cut a release.
Development
pnpm install
pnpm rebuild esbuild # pnpm 11 blocks esbuild's postinstall by default
pnpm build # build all packages
pnpm test # run all tests
pnpm typecheck # type-check all packages
pnpm lint # prettier check
pnpm smoke # verify the built server exposes only read-only tools
License
MIT. See LICENSE.
Disclaimer
This is an independent community project. It is not affiliated with, sponsored by, maintained by, or endorsed by Helcim Inc. "Helcim" is a trademark of Helcim Inc. and is used here only to describe API compatibility. This project does not use Helcim logos or branding.
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
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.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.