AICre8 MCP Server

Instructions

Implementation Reference

• src/client.ts:73-113 (handler)

  The generateCode method in AICre8Client class handles the core logic: makes a POST request to /projects/{projectId}/generate endpoint, handles SSE stream response by reading chunks and returning the full accumulated result text.

  async generateCode(
    projectId: string,
    params: { prompt: string; chat_mode?: 'build' | 'discuss' },
  ): Promise<string> {
    const url = `${this.baseUrl}/projects/${projectId}/generate`;
  
    const res = await fetch(url, {
      method: 'POST',
      headers: this.headers,
      body: JSON.stringify(params),
    });
  
    if (!res.ok) {
      const error = await res.json().catch(() => ({ error: res.statusText }));
      throw new Error(
        `AICre8 API error ${res.status}: ${(error as any).error || res.statusText}`,
      );
    }
  
    // The response is an SSE stream — consume it and return the full text
    const reader = res.body?.getReader();
  
    if (!reader) {
      throw new Error('No response body');
    }
  
    const decoder = new TextDecoder();
    let result = '';
  
    while (true) {
      const { done, value } = await reader.read();
  
      if (done) {
        break;
      }
  
      result += decoder.decode(value, { stream: true });
    }
  
    return result;
  }

• src/index.ts:86-112 (registration)

  Tool registration with MCP server using server.tool(). Registers 'generate_code' with description, Zod schema for inputs (project_id, prompt, chat_mode), and async handler that calls client.generateCode() and returns the result as text content.

  // ── Tool: generate_code ──
  
  server.tool(
    'generate_code',
    'Send a prompt to generate or modify code in an AICre8 project using AI. The AI will write files, install packages, and update the project. Costs 1 credit per call.',
    {
      project_id: z.string().describe('Project ID (UUID or url_id)'),
      prompt: z.string().describe('What to build or change'),
      chat_mode: z
        .enum(['build', 'discuss'])
        .optional()
        .describe('Mode: "build" generates code (default), "discuss" for conversation only'),
    },
    async (params) => {
      try {
        const result = await client.generateCode(params.project_id, {
          prompt: params.prompt,
          chat_mode: params.chat_mode,
        });
        return {
          content: [{ type: 'text' as const, text: result }],
        };
      } catch (err: any) {
        return { content: [{ type: 'text' as const, text: `Error: ${err.message}` }], isError: true };
      }
    },
  );

• src/index.ts:91-98 (schema)

  Zod schema defining the input parameters for generate_code tool: project_id (required string), prompt (required string), and chat_mode (optional enum of 'build' or 'discuss').

  {
    project_id: z.string().describe('Project ID (UUID or url_id)'),
    prompt: z.string().describe('What to build or change'),
    chat_mode: z
      .enum(['build', 'discuss'])
      .optional()
      .describe('Mode: "build" generates code (default), "discuss" for conversation only'),
  },