# KBBI [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/gaato/kbbi-mcp  
**GitHub Stars:** 1  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/kbbi

## Description
Look up Indonesian words in KBBI, the official Indonesian dictionary (unofficial client).

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

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

## Documentation

## What the KBBI MCP server does

The KBBI MCP server gives MCP-compatible AI assistants read-only access to KBBI Daring, the official Indonesian dictionary. It accepts a word or phrase and returns structured dictionary data rather than only page text. Results can include multiple homographs, pronunciation, root words, word classes, definitions, and example sentences with their meanings.

This is an unofficial client and is not affiliated with or endorsed by the Language Development and Cultivation Agency or KBBI Daring. The project states that dictionary content belongs to its copyright holders and that the server is intended for personal, non-commercial lookups.

## How it works

The server runs locally and communicates with the MCP client over standard input and output. For each lookup, it fetches the corresponding KBBI entry page at the configured site, parses the page, and returns the extracted structure. An equivalent JSON result is available through a `kbbi://{query}` resource.

The `kbbi_lookup` tool requires one string parameter named `query`. It accepts either a word, such as `makan`, or a phrase, such as `rumah sakit`. Empty queries and failed requests, including cases where KBBI cannot be reached, are reported as tool errors. Results are cached in memory for the lifetime of the running process.

## Setup and configuration

The standard configuration starts the package through `uvx`:

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

Install `uv` first. If `uv` is unavailable, install the package with `pip install kbbi-mcp` and use `kbbi-mcp` or `python -m kbbi_mcp` as the command. The README provides setup commands or configuration guidance for Claude Code, Codex CLI, Gemini CLI, VS Code, Cursor, and Claude Desktop.

The KBBI MCP server does not require an API key or account. Optional environment variables control request timeout with `KBBI_TIMEOUT_SECONDS`, logging verbosity with `KBBI_LOG_LEVEL`, and the dictionary endpoint with `KBBI_BASE_URL`. The default endpoint is `https://kbbi.kemendikdasmen.go.id` and the default timeout is 10 seconds.

## Tools and capabilities

- `kbbi_lookup`: Performs a read-only lookup for an Indonesian word or phrase.
- `kbbi://{query}`: Provides the same lookup result as JSON through an MCP resource.
- Structured output: Returns headwords, homograph numbers, pronunciation, roots, word classes, definitions, examples, and example meanings when present.

These results can support agent responses that explain Indonesian vocabulary, identify a root word, interpret a phrase, or quote dictionary definitions.

## Limitations and notes

Anonymous lookups do not include information KBBI shows only to signed-in users, such as etymology, related entries, or suggestions for missing words. If a lookup returns nothing, try the base word or a different spelling.

KBBI limits anonymous searches, so avoid rapid bulk requests. Each lookup depends on KBBI’s current page markup; changes to that markup can cause parsing failures until the package is updated. The server is therefore better suited to occasional interactive lookups than uncontrolled high-volume collection. Use the MCP Inspector with `npx @modelcontextprotocol/inspector uvx kbbi-mcp` when diagnosing client or server behavior.

_Full upstream README: https://allmcps.com/mcp/kbbi/readme_