kbbi-mcp

An MCP server for querying KBBI (Kamus Besar Bahasa Indonesia / KBBI Daring).

Python: 3.13+

This project exposes a single, stable JSON tool output so LLM clients can decide how to format, translate, or summarize results.

Relationship to KBBI Daring

This project is unofficial and is not affiliated with or endorsed by the official KBBI Daring service.

Features

• MCP tool: kbbi_lookup(query: str)
• MCP resource: kbbi://{query} (same payload as kbbi_lookup)
• No login/auth flow required
• Uses official KBBI VI Daring by default

Configure in an MCP client (JSON)

Most MCP clients (including Claude Desktop) use a JSON config with a top-level mcpServers object. This mcpServers format is an emergent standard across the MCP ecosystem (see: https://gofastmcp.com/integrations/mcp-json-configuration.md).

mcpServers-based clients (Claude Desktop / Cursor / Windsurf)

This matches the convention used by many Python MCP servers.

Where to put it:

• Claude Desktop: ~/.claude/claude_desktop_config.json
• Cursor: .cursor/mcp.json (project) or ~/.cursor/mcp.json (global)
• Windsurf: ~/.codeium/windsurf/mcp_config.json

{
	"mcpServers": {
		"kbbi": {
			"command": "uvx",
			"args": ["kbbi-mcp"]
		}
	}
}

Note: this example uses uvx (part of uv) to run the server. Install uv here: https://docs.astral.sh/uv/getting-started/installation/

If you don't want to depend on uv, install kbbi-mcp into an environment and point your client to that environment's executable. For example:

• Use the console script (recommended when available): kbbi-mcp
• Or run the module: python -m kbbi_mcp

The server performs direct lookups to the official KBBI VI Daring site.

Local development (run from this repo)

You'll need uv installed: https://docs.astral.sh/uv/getting-started/installation/

This repo includes a fastmcp.json file that defines how to run the server (source + uv environment + stdio transport).

To run the server directly from this checkout:

fastmcp run

To generate an mcpServers entry you can paste into your MCP client config:

fastmcp install mcp-json fastmcp.json

Note: the generated configuration uses absolute paths so it works regardless of the client's working directory.

Tool: kbbi_lookup

Input

• query (string): a word or phrase

Example tool arguments:

{
	"query": "makan"
}

Output

Returns a JSON object:

• found (bool)
• query (string)
• url (string | null)
• entries (list)
• suggestions (list)
• error (string, optional): present only when the request is invalid (e.g. empty query) or an unexpected error occurs

entries uses an English-key schema with normalization:

• etymology is always present (as an object or null)
• related-word lists are always present (as arrays, possibly empty)

This keeps tool output predictable across source variations.

Example tool output:

{
	"found": true,
	"query": "makan",
	"url": "https://kbbi.kemendikdasmen.go.id/entri/makan",
	"entries": [
		{
			"headword": "makan",
			"sense_number": "",
			"root_words": [],
			"pronunciation": "",
			"nonstandard_forms": [],
			"variants": [],
			"definitions": [],
			"etymology": null,
			"derived_words": [],
			"compound_words": [],
			"proverbs": [],
			"idioms": []
		}
	],
	"suggestions": []
}

Resource: kbbi://{query}

This server also exposes the same payload as a read-only MCP resource.

• kbbi://makan

For low-level debugging, a client would read it using resources/read with {"uri": "kbbi://makan"}.

Data source

Lookup behavior:

• Official source: https://kbbi.kemendikdasmen.go.id/entri/{query}

Optional environment variables:

• KBBI_BASE_URL (default: https://kbbi.kemendikdasmen.go.id)
• KBBI_TIMEOUT_SECONDS (default: 10.0)