Gemini MCP Server for Claude Desktop

Instructions

Implementation Reference

• src/tools/image-editing.js:55-137 (handler)

  The main handler function that performs image editing: validates inputs, loads and encodes image, enhances edit instruction using intelligence system, calls Gemini API for editing, saves the output image, learns from interaction, and returns formatted response.
  async execute(args) { const imagePath = validateNonEmptyString(args.image_path, 'image_path'); const editInstruction = validateNonEmptyString(args.edit_instruction, 'edit_instruction'); const context = args.context ? validateString(args.context, 'context') : null; log(`Editing image: "${imagePath}" with instruction: "${editInstruction}" and context: ${context || 'general'}`, this.name); try { validateFileSize(imagePath, config.MAX_IMAGE_SIZE_MB); const imageBuffer = readFileAsBuffer(imagePath); const imageBase64 = imageBuffer.toString('base64'); const mimeType = getMimeType(imagePath, config.SUPPORTED_IMAGE_MIMES); log(`Image file loaded: ${(imageBuffer.length / 1024).toFixed(2)}KB, MIME type: ${mimeType}`, this.name); let enhancedEditInstruction = editInstruction; if (this.intelligenceSystem.initialized) { try { enhancedEditInstruction = await this.intelligenceSystem.enhancePrompt(editInstruction, context, this.name); log('Applied Tool Intelligence enhancement', this.name); } catch (err) { log(`Tool Intelligence enhancement failed: ${err.message}`, this.name); } } let editPrompt = `Please edit this image according to the following instruction: ${enhancedEditInstruction}`; // eslint-disable-line max-len if (context) { editPrompt += `\n\nAdditional context: ${context}`; } editPrompt += '\n\nProvide the edited image as output. Maintain the overall composition and quality while making the requested changes.'; // eslint-disable-line max-len const editedImageData = await this.geminiService.analyzeImage('IMAGE_EDITING', editPrompt, imageBase64, mimeType); if (editedImageData) { log('Successfully extracted edited image data', this.name); ensureDirectoryExists(config.OUTPUT_DIR, this.name); const timestamp = Date.now(); const hash = crypto.createHash('md5').update(editInstruction).digest('hex'); const imageName = `gemini-edited-${hash}-${timestamp}.png`; const outputImagePath = path.join(config.OUTPUT_DIR, imageName); fs.writeFileSync(outputImagePath, Buffer.from(editedImageData, 'base64')); log(`Edited image saved to: ${outputImagePath}`, this.name); if (this.intelligenceSystem.initialized) { try { await this.intelligenceSystem.learnFromInteraction(editInstruction, enhancedEditInstruction, `Image edited successfully: ${outputImagePath}`, context, this.name); // eslint-disable-line max-len log('Tool Intelligence learned from interaction', this.name); } catch (err) { log(`Tool Intelligence learning failed: ${err.message}`, this.name); } } let finalResponse = `✓ Image successfully edited with instruction: "${editInstruction}"\n\n**Original:** ${imagePath}\n**Edited:** ${outputImagePath}\n\n**Edit Applied:** ${editInstruction}${context ? `\n**Context:** ${context}` : ''}`; // eslint-disable-line max-len if (context && this.intelligenceSystem.initialized) { finalResponse += `\n\n---\n_Enhancement applied based on context: ${context}_`; } return { content: [ { type: 'text', text: finalResponse, }, ], }; } log('No edited image data found in response', this.name); return { content: [ { type: 'text', text: `Could not edit image with instruction: "${editInstruction}". No edited image data was returned by Gemini API. Try rephrasing your edit instruction or using a different image.`, // eslint-disable-line max-len }, ], }; } catch (error) { log(`Error editing image: ${error.message}`, this.name); throw new Error(`Error editing image: ${error.message}`); } }
• src/tools/image-editing.js:24-41 (schema)

  JSON schema defining the input parameters for the tool: image_path (required), edit_instruction (required), and optional context.
  { type: 'object', properties: { image_path: { type: 'string', description: 'Path to the image file to edit (JPEG, PNG, WebP, GIF, BMP)', }, edit_instruction: { type: 'string', description: 'Detailed instruction for how to edit the image', }, context: { type: 'string', description: 'Optional context for intelligent enhancement (e.g., "subtle", "dramatic", "professional")', }, }, required: ['image_path', 'edit_instruction'], },
• src/tools/index.js:79-79 (registration)

  Registers the ImageEditingTool instance (named 'gemini-edit-image') into the tools registry.
  registerTool(new ImageEditingTool(intelligenceSystem, geminiService));