Streamline MCP Server

An MCP (Model Context Protocol) server that exposes the Streamline REST API as tools for AI agents. It allows AI assistants like Claude and Cursor to interact with listings, reservations, guests, work orders, and other Streamline resources through natural language.

Quick Start

npx @streamlinevrs/mcp-server

Or install globally:

npm install -g @streamlinevrs/mcp-server
streamline-mcp

Authentication

The MCP server runs locally over stdio (no network listener). It authenticates to the Streamline API using your existing M2M client credentials — the same client_id and client_secret you use for direct API access.

Tokens are exchanged automatically via POST /v1/auth/token and refreshed before expiry. No manual JWT management needed.

| Variable | Description | Required | |---|---|---| | STL_API_URL | Streamline API base URL (default: https://api-staging.streamlinevrs.com) | No | | STL_CLIENT_ID | Your Auth0 M2M client ID | Yes | | STL_CLIENT_SECRET | Your Auth0 M2M client secret | Yes | | STL_COMPANY_CONTEXT | Tenant company id, sent as the X-Company-Context header on every request (see Tenant scope) | Partner/client: yes · Staff: no |

A static STL_API_TOKEN is also supported as a fallback (if set, it takes priority over client credentials).

To obtain credentials, contact your Streamline administrator.

Tenant scope (X-Company-Context)

Most tools (listings, reservations, guests, users, …) are company-scoped, so the API needs to know which tenant you're acting on. The server reads STL_COMPANY_CONTEXT and sends it as the X-Company-Context header on every request. What you need to set depends on the credentials you were issued:

| Your credentials | STL_COMPANY_CONTEXT | Per-call override on api_request | |---|---|---| | Partner / client (scoped to specific companies) | Required — only your assigned companies | Reads (GET) only | | Streamline staff (can reach any company) | Optional | Reads (GET) only |

Partner / client credentials — set it (required)

Partner credentials carry no implicit company and may only act on companies assigned to them in Auth0. Set STL_COMPANY_CONTEXT to the company id you manage. Without it, every company-scoped tool returns 403 Company context required; targeting a company you aren't assigned to also returns 403. To work across several assigned companies, add one server entry per company, each with its own STL_COMPANY_CONTEXT.

Streamline staff credentials — optional, with a read override

Staff can reach any company but also carry no implicit one, so STL_COMPANY_CONTEXT is optional:

• Leave it unset to use the root tools (list_companies, list_domains, health_check) and pick a tenant per read.
• Set it to pin a default tenant for the whole session.
• Override a single read with the api_request tool's company_context argument — e.g. inspect another company without restarting.

Writes are always pinned (everyone)

The per-call company_context override is accepted on GET only. POST/PATCH/PUT/DELETE always use STL_COMPANY_CONTEXT, and passing an override on a write is rejected. This guarantees a drifted or hallucinated id can never redirect a mutation to the wrong (but still authorized) tenant — to write to a tenant, that company must be the configured STL_COMPANY_CONTEXT.

Client Configuration

Automatic Setup (Recommended)

Run the interactive setup command — it detects the correct npx path and writes the config file for you:

npx -y @streamlinevrs/mcp-server --setup

You'll be prompted for your API URL, client ID, client secret, and an optional company context (see Tenant scope — required for partner/client credentials). The command supports both Cursor and Claude Desktop.

After setup, restart your client to activate the MCP server.

Manual Configuration

If the setup command doesn't work for your environment, configure manually. First, find your full npx path:

which npx    # macOS/Linux
where npx    # Windows

Use the full path (e.g., /opt/homebrew/bin/npx) as the command value, and include a PATH entry pointing to the directory containing node — most MCP clients don't inherit your shell's PATH, so both npx and node will fail without these.

Cursor

Add to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "streamline": {
      "command": "/opt/homebrew/bin/npx",
      "args": ["-y", "@streamlinevrs/mcp-server"],
      "env": {
        "PATH": "/opt/homebrew/bin:/usr/local/bin:/usr/bin:/bin",
        "STL_API_URL": "https://api-staging.streamlinevrs.com",
        "STL_CLIENT_ID": "your-client-id",
        "STL_CLIENT_SECRET": "your-client-secret",
        "STL_COMPANY_CONTEXT": "54"
      }
    }
  }
}

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS):

{
  "mcpServers": {
    "streamline": {
      "command": "/opt/homebrew/bin/npx",
      "args": ["-y", "@streamlinevrs/mcp-server"],
      "env": {
        "PATH": "/opt/homebrew/bin:/usr/local/bin:/usr/bin:/bin",
        "STL_API_URL": "https://api-staging.streamlinevrs.com",
        "STL_CLIENT_ID": "your-client-id",
        "STL_CLIENT_SECRET": "your-client-secret",
        "STL_COMPANY_CONTEXT": "54"
      }
    }
  }
}

Replace /opt/homebrew/bin with the directory from which npx on your machine. STL_COMPANY_CONTEXT is required for partner credentials and optional for staff — omit the line if you don't want a default tenant scope.

Available Tools

The server uses lazy-loading to keep the initial tool list small. Only 4 tools are registered at startup:

| Tool | Description | |---|---| | list_domains | Lists all available API domains with endpoint counts and activation status | | activate_domain | Loads all tools for a specific domain (e.g. "Listings", "Reservations") | | api_request | Generic fallback — make any API call by specifying method, path, body, and query params | | health_check | Quick connectivity check against the API |

When an AI calls activate_domain, the domain's tools are dynamically registered and the client is notified via tools/list_changed. This keeps tool selection fast and focused.

Usage Pattern

1. Call list_domains to see all available domains
2. Call activate_domain with the domain you need (e.g. "Listings")
3. The domain's tools appear and are ready to use
4. Use api_request as a fallback for any endpoint without activating its domain

Available Domains

Domains are auto-generated from the Streamline API OpenAPI specification:

| Domain | Endpoints | |---|---| | Accounting | 6 | | Airbnb Integration | 5 | | Authentication | 1 | | Channel Listings | 3 | | Channels | 5 | | Companies | 7 | | Company Assets | 1 | | Company Data | 4 | | Emails | 4 | | Features | 3 | | Folios | 2 | | Guests | 6 | | Housekeeping Tasks | 7 | | Inbox | 3 | | Listings | 26 | | Owners | 9 | | Payments | 2 | | Reservation Flags | 3 | | Reservations | 9 | | Reviews | 3 | | Settings | 1 | | Statistics | 1 | | Taxes | 4 | | User Data | 4 | | User Groups | 5 | | Users | 8 | | Work Orders | 9 |

A read-only resource is also available: api-docs (docs://openapi) returns the full OpenAPI specification.