/**
* Zen MCP Wrapper Template
*
* ⚠️ COPY THIS FILE TO YOUR PROJECT - DO NOT IMPORT FROM THIS PACKAGE
*
* Why? MCP servers update independently. This template shows the current
* parameter schema, but zen-mcp-server may change at any time.
*
* Usage:
* 1. Copy utils-template.ts to src/lib/mcp/utils.ts
* 2. Copy this file to src/lib/mcp/zen.ts
* 3. Install zen-mcp-server in your .mcp.json
* 4. Adapt parameters to match YOUR installed zen version
* 5. Maintain it when zen-mcp-server updates
*
* Last verified: 2024-11-09 with zen-mcp-server from BeehiveInnovations
*/
import { callMCPToolSafe, parseMCPResult, parseStringResult } from './utils';
/**
* Available models for zen tools
*/
export type ZenModel =
| 'gemini-2.5-pro' // Recommended for deep thinking
| 'gemini-2.5-flash' // Faster, less thorough
| 'o3-pro' // OpenAI O3
| 'o3-mini'
| string;
/**
* Deep reasoning with step tracking
*
* ⚠️ Parameters verified as of 2024-11-09
* Check zen-mcp-server docs if this breaks after updates
*/
export async function zenThinkDeep(
question: string,
options: {
model?: ZenModel;
steps?: number;
} = {}
): Promise<{
status: 'pause_for_thinkdeep' | 'complete';
content?: string;
findings?: string;
continuation_id?: string;
thinkdeep_status?: {
current_confidence: 'low' | 'medium' | 'high';
};
}> {
const { model = 'gemini-2.5-pro', steps = 1 } = options;
const result = await callMCPToolSafe(
'mcp__zen__thinkdeep',
{
step: question, // Current param name (was 'query' before)
step_number: 1,
total_steps: steps,
next_step_required: steps > 1,
findings: '', // Current: string (was array before)
model // Current param name (was 'cli_name' before)
},
'zen thinkdeep'
);
return parseMCPResult(result);
}
/**
* Code review with findings tracking
*/
export async function zenCodeReview(
code: string,
language: string,
model: ZenModel = 'gemini-2.5-pro'
): Promise<{
content: string;
findings: any[];
}> {
const result = await callMCPToolSafe(
'mcp__zen__codereview',
{
code,
language,
findings: [],
step: 'security-and-quality-review',
step_number: 1,
total_steps: 1,
next_step_required: false,
model
},
'zen codereview'
);
return parseMCPResult(result);
}
/**
* Direct CLI execution
*/
export async function zenClink(
prompt: string,
persona: string = 'assistant',
model: ZenModel = 'gemini-2.5-pro'
): Promise<string> {
const result = await callMCPToolSafe(
'mcp__zen__clink',
{
prompt,
persona,
model
},
'zen clink'
);
return parseStringResult(result);
}
/**
* List available models
*/
export async function zenListModels(): Promise<{
gemini: string[];
openai: string[];
}> {
const result = await callMCPToolSafe(
'mcp__zen__listmodels',
{},
'zen listmodels'
);
return parseMCPResult(result);
}
// Add more zen tool wrappers as needed...
