A TypeScript-based MCP server providing data sources, agents, and conversation tooling via JSON-RPC.
Configuration
View docs{
"mcpServers": {
"ma3u-dust-mcp-server-postman-railway": {
"command": "node",
"args": [
"mcpServer.js"
],
"env": {
"DUST_API_KEY": "YOUR_API_KEY",
"DUST_AGENT_ID": "YOUR_AGENT_ID",
"DUST_WORKSPACE_ID": "YOUR_WORKSPACE_ID"
}
}
}
}You are running a TypeScript-based MCP server that enables structured tool access and conversation handling for agents. It powers REST-like and JSON-RPC interactions, supports hot-reloading during development, and is designed to be easily configured for workspace-specific APIs and agents.
Start by running the MCP server locally and connecting an MCP client of your choice. The server exposes a set of tools you can call via JSON-RPC and manages conversations, sessions, and history so you can build interactive agent experiences. You will typically start a new session, create a conversation, and then send messages that the server forwards to the agent or external Dust services. You can also run with server-sent events support for real-time updates if your client requires it.
Prerequisites: ensure you have Node.js 18+ (Node.js 20+ recommended) and npm installed. You will also want Git for cloning repositories and TypeScript for development.
Step-by-step commands to set up and run the server from a project directory.
Configuration notes: environment-based configuration is used. Create a .env file in the project root and populate required values such as tokens or workspace IDs as shown in the variables section.
Troubleshooting: common issues include the server not starting, connection timeouts, or invalid method errors. Verify Node.js is in your PATH, dependencies are installed, and environment variables are correctly loaded. Check server logs for detailed error messages.
List all workspace vaults available in the MCP environment
List available assistants that can be invoked by the MCP server
List data source views to inspect available data representations
Retrieve events for a given conversation to audit or replay interactions
Fetch available data sources configured for the workspace
Search for assistants by name to discover matching capabilities
Get details about a specific conversation including messages and state
Retrieve a document by ID from the document store
Get details about an application run for traceability
Get event history for a specific message in a conversation
Create or update a document in the repository
Retrieve multiple documents by IDs or queries
Start a new conversation session
Send a message in an active conversation
Create a content fragment to accompany a message or document
Start a new application run to execute a workflow or task
Search within a data source for matching records
Search within a data source view for specific entries