MCP server
Connect SoftSolz to Cursor, VS Code, Claude and any MCP client.
The @softsolz/mcp package is the official Model Context Protocol server for SoftSolz. Point an AI tool at it and the model can list your forms, create invoices, inspect submissions, run workflows and more - every call scoped to your API key.
No extension to install, and no server URL to copy
This is a local stdio server, not a hosted endpoint - it runs on your machine via npx and talks to https://app.softsolz.uk for you. There is nothing to deploy and no URL to paste. You do notinstall anything from the VS Code or Cursor marketplace either: MCP support is built in (VS Code needs GitHub Copilot's agent mode; Cursor has it natively). You just add the config below.
What you need
- Node.js 18.17+ (so
npxis available). - A SoftSolz API key. Create one under Developers → API keys. Start with an
sk_test_*key so every experiment lands in your sandbox workspace.
The command every client runs
This single command is the whole server. Each editor just stores it in config and launches it for you:
SOFTSOLZ_API_KEY=sk_test_your_key npx -y @softsolz/mcpVS Code
One-click install, or add it manually.
Or run the CLI (then replace the key, or switch to an input variable as below):
code --add-mcp "{\"name\":\"softsolz\",\"command\":\"npx\",\"args\":[\"-y\",\"@softsolz/mcp\"],\"env\":{\"SOFTSOLZ_API_KEY\":\"sk_test_your_key\"}}"Or commit it to your workspace at .vscode/mcp.json. Using an input keeps the key out of the file - VS Code prompts for it on first run:
{ "servers": { "softsolz": { "command": "npx", "args": ["-y", "@softsolz/mcp"], "env": { "SOFTSOLZ_API_KEY": "${input:softsolz-api-key}" } } }, "inputs": [ { "type": "promptString", "id": "softsolz-api-key", "description": "SoftSolz API key (sk_test_* or sk_live_*)", "password": true } ]}Then open the Copilot Chat Agent mode and the SoftSolz tools appear in the tool picker.
Cursor
One-click install, or add it manually.
After installing, open ~/.cursor/mcp.json (or .cursor/mcp.json in your project) and set your real key:
{ "mcpServers": { "softsolz": { "command": "npx", "args": ["-y", "@softsolz/mcp"], "env": { "SOFTSOLZ_API_KEY": "sk_test_your_key" } } }}Claude Code
claude mcp add softsolz -e SOFTSOLZ_API_KEY=sk_test_your_key -- npx -y @softsolz/mcpClaude Desktop
Add this to claude_desktop_config.json (Settings → Developer → Edit Config):
{ "mcpServers": { "softsolz": { "command": "npx", "args": ["-y", "@softsolz/mcp"], "env": { "SOFTSOLZ_API_KEY": "sk_test_your_key" } } }}Configuration
| Env var | Required | Description |
|---|---|---|
SOFTSOLZ_API_KEY | Yes | Secret key. sk_test_* routes to your sandbox workspace automatically. |
SOFTSOLZ_BASE_URL | No | API origin override (default https://app.softsolz.uk). |
SOFTSOLZ_SERVICES | No | Comma-separated filter, e.g. forms,invoicing. Fewer tools means sharper AI tool selection. |
Start in the sandbox
Use an sk_test_*key first so the AI's actions stay in your sandbox workspace. Switch the env var to sk_live_* only once you trust the setup. The server can never exceed the scopes on the key - a key with only service.forms.view can list forms but not create invoices.
What it can do
You get two platform tools (whoami, list_services) plus one tool per API operation across every SoftSolz service - around 120 tools, named <service>_<operation> (for example forms_list_forms, invoicing_create_invoice). Browse the underlying endpoints in the API Reference.