# Clawdbot MCP Plugin - Installation Guide
## Correct Configuration Method
According to Clawdbot documentation, plugins should be configured in `clawdbot.json`, NOT in `config.yaml`.
### Plugin Manifest
The plugin defines its configuration schema in `clawdbot.plugin.json`:
```json
{
"id": "clawdbot-mcp-plugin",
"name": "MCP Server Bridge",
"description": "Enables Model Context Protocol (MCP) server integration with progressive tool disclosure",
"version": "0.1.0",
"configSchema": {
"type": "object",
"additionalProperties": false,
"properties": {
"configFilePath": {
"type": "string",
"description": "Path to mcp-config.yaml file (supports one-mcp's YAML format)",
"default": ".clawdbot/mcp-config.yaml"
},
"serverId": {
"type": "string",
"description": "Unique identifier for the toolkit",
"default": "clawdbot-mcp"
},
"noCache": {
"type": "boolean",
"description": "Disable configuration caching",
"default": false
}
}
}
}
```
### Installation Steps
1. **Install the plugin** using the Clawdbot CLI:
```bash
cd /Users/vuongngo/workspace/aicode-toolkit/packages/clawdbot-mcp-plugin
pnpm build
clawdbot plugins install <path-or-method>
```
2. **Configure in clawdbot.json** (`~/.clawdbot/clawdbot.json`):
After installation, add configuration to the `plugins.entries` section.
**CRITICAL**: Config values must be nested under a `"config"` key:
```json
{
"plugins": {
"entries": {
"clawdbot-mcp-plugin": {
"enabled": true,
"config": {
"configFilePath": "/Users/vuongngo/.clawdbot/mcp-config.yaml",
"serverId": "clawdbot-toolkit",
"noCache": false
}
}
}
},
"agents": {
"list": [
{
"id": "main",
"tools": {
"allow": [
"mcp__describe_tools",
"mcp__use_tool"
]
}
}
]
}
}
```
### MCP Server Configuration
Create `~/.clawdbot/mcp-config.yaml`:
```yaml
mcpServers:
memory:
name: "Memory Storage"
instruction: "Key-value storage"
transport: "stdio"
command: "npx"
args: ["-y", "@modelcontextprotocol/server-memory"]
filesystem:
name: "Filesystem Operations"
instruction: "File operations"
transport: "stdio"
command: "npx"
args: ["-y", "@modelcontextprotocol/server-filesystem", "/Users/vuongngo/workspace"]
```
**Note**: `command` and `args` are at top level, NOT nested under `config`.
### How Plugin Configuration Works
1. Plugin exports an **object** with `id`, `name`, `description`, `configSchema`, and `register()` method
2. Clawdbot reads the plugin manifest (`clawdbot.plugin.json`) during plugin discovery
3. Configuration is stored in `clawdbot.json` under `plugins.entries[pluginId].config` (nested!)
4. Plugin accesses config via `api.pluginConfig` (NOT `api.getConfig()`)
5. Configuration must match the schema defined in both the manifest and plugin code
### Key Points
- ❌ Do NOT use `config.yaml` for plugin configuration
- ✅ Use `clawdbot.json` for plugin configuration
- Plugin must be installed/discovered before adding configuration
- Configuration keys must match the `configSchema` in the manifest
- Plugin ID in config must match the `id` in `clawdbot.plugin.json`
### Verification
Check plugin status:
```bash
clawdbot plugins list | grep -A 3 "MCP Server"
```
Expected output:
```
│ MCP Server │ clawdbot │ loaded │ ...
│ Bridge │ -mcp- │ │ ...
│ │ plugin │ │ ...
```
## TODO
- Document the exact installation command that works
- Verify plugin configuration is recognized after installation
- Test MCP server connections when gateway starts
