feat: add save_image tool and list image MCP tools

Expose save_image across direct model, MCP gateway, DeepSeek Harness and remote Agents (written on the remote host via chunked reads), and show the image MCP tool list in settings.

src/renderer/src/McpSettingsSection.test.tsx

@@ -148,6 +148,22 @@ it('preserves the Obsidian draft after failures and folder-picker cancellation',
   })
 })
 
+it('lists the image generation and save tools on the managed image card', async () => {
+  const getSnapshot = vi.fn().mockResolvedValue({ mcpServers: [], skills: [] })
+  vi.stubGlobal('goodbuddy', { capabilities: { getSnapshot } })
+  render(<McpSettingsSection onOpenImageModelSettings={vi.fn()} />)
+  await waitFor(() => expect(getSnapshot).toHaveBeenCalledOnce())
+  const name = i18n.t('integrations:conversationImages.title')
+  const toggle = screen.getByRole('button', { name: i18n.t('integrations:mcp.builtin.expandServer', { name }) })
+  expect(toggle).toHaveTextContent(i18n.t('integrations:mcp.builtin.toolCount', { count: 2 }))
+  expect(screen.queryByText('save_image')).not.toBeInTheDocument()
+  fireEvent.click(toggle)
+  const tools = screen.getByRole('region', { name: i18n.t('integrations:mcp.builtin.toolsAriaLabel', { name }) })
+  expect(tools).toHaveTextContent('generate_image')
+  expect(tools).toHaveTextContent('save_image')
+  expect(tools).toHaveTextContent(i18n.t('integrations:conversationImages.tools.save_image'))
+})
+
 it('removes the redundant MCP title and places the custom action in content', async () => {
   const getSnapshot = vi.fn().mockResolvedValue({ mcpServers: [], skills: [] })
   vi.stubGlobal('goodbuddy', { capabilities: { getSnapshot } })

src/renderer/src/McpSettingsSection.tsx

@@ -16,7 +16,10 @@ import { useEffect, useMemo, useRef, useState } from 'react'
 import { createPortal } from 'react-dom'
 import { useTranslation } from 'react-i18next'
 import { builtinMcpServers } from '../../shared/builtin-mcp-servers'
-import { builtinModelToolGroups } from '../../shared/builtin-model-tools'
+import {
+  builtinModelToolGroups,
+  builtinModelTools
+} from '../../shared/builtin-model-tools'
 import type {
   BuiltinMcpServerId,
   CapabilityDiagnosticReport,
@@ -40,6 +43,8 @@ const configurableMcpTargets: RuntimeTarget[] = [
   'continue',
   'deepseek-harness'
 ]
+const imageTools = builtinModelTools.filter((tool) => tool.group === 'image')
+const imageExpansionId = 'builtin:image-generation'
 type McpSettingsTab =
   | 'builtin'
   | 'custom'
@@ -647,6 +652,65 @@ export function McpSettingsSection({
             >
               {t('conversationImages.openSettings')}
             </button>
+            <button
+              aria-controls="mcp-server-tools-image-generation"
+              aria-expanded={expandedItemIds.has(imageExpansionId)}
+              aria-label={t(
+                expandedItemIds.has(imageExpansionId)
+                  ? 'mcp.builtin.collapseServer'
+                  : 'mcp.builtin.expandServer',
+                { name: t('conversationImages.title') }
+              )}
+              className="secondary-button builtin-mcp-card__details"
+              onClick={() => toggleItem(imageExpansionId)}
+              type="button"
+            >
+              {t('mcp.builtin.toolCount', { count: imageTools.length })}
+              <ChevronDown
+                aria-hidden="true"
+                className={
+                  expandedItemIds.has(imageExpansionId)
+                    ? 'mcp-server-card__chevron mcp-server-card__chevron--expanded'
+                    : 'mcp-server-card__chevron'
+                }
+                size={15}
+              />
+            </button>
+            {expandedItemIds.has(imageExpansionId) && (
+              <div
+                className="builtin-mcp-card__tools"
+                id="mcp-server-tools-image-generation"
+              >
+                <section
+                  aria-label={t('mcp.builtin.toolsAriaLabel', {
+                    name: t('conversationImages.title')
+                  })}
+                  className="mcp-server-tools"
+                >
+                  <div className="mcp-server-tools__heading">
+                    <strong>{t('mcp.builtin.tools')}</strong>
+                    <small>
+                      {t('mcp.profiles.count', { count: imageTools.length })}
+                    </small>
+                  </div>
+                  <ul>
+                    {imageTools.map((tool) => (
+                      <li key={tool.name}>
+                        <div>
+                          <code>{tool.name}</code>
+                          <span className="builtin-tool-badge">
+                            {tool.access === 'write'
+                              ? t('mcp.builtin.write')
+                              : t('mcp.builtin.readOnly')}
+                          </span>
+                        </div>
+                        <p>{t(`conversationImages.tools.${tool.name}`)}</p>
+                      </li>
+                    ))}
+                  </ul>
+                </section>
+              </div>
+            )}
           </article>
           {builtinMcpServers.map((server) => {
             const expansionId = `builtin:${server.id}`

src/renderer/src/i18n/locales/en-US/integrations.ts

@@ -7,7 +7,11 @@ export const integrations = {
     managed: 'Builtin · Automatically managed',
     description: 'Automatically managed by the image models’ “Allow AI to invoke in conversations” setting.',
     runtimeScope: 'Assignments include applicable remote Runtimes through GoodBuddy Agent. Generation requires Execute and a chat model with tool calling; assignments do not indicate runtime readiness.',
-    openSettings: 'Go to image model settings'
+    openSettings: 'Go to image model settings',
+    tools: {
+      generate_image: 'Generate a new image with an image model allowed in conversations, or edit images from this conversation. Results are saved in the conversation.',
+      save_image: 'Save a generated or uploaded image from this conversation to a PNG, JPEG or WebP file. Requires Execute mode; in remote projects the file is written on the remote host.'
+    }
   },
   channels: {
     tabs: {

src/renderer/src/i18n/locales/zh-CN/integrations.ts

@@ -4,7 +4,11 @@ export const integrations = {
     managed: '内置 · 自动管理',
     description: '由图片模型的「允许 AI 在会话中调用」设置自动管理。',
     runtimeScope: '分配包含 GoodBuddy Agent 远程路径上的适用 Runtime。生成需要 Execute 模式及支持工具调用的聊天模型；勾选不代表 Runtime 已就绪。',
-    openSettings: '前往图片模型设置'
+    openSettings: '前往图片模型设置',
+    tools: {
+      generate_image: '使用已允许会话调用的图片模型生成新图片，或基于会话中的图片进行编辑，结果保存在当前会话。',
+      save_image: '将当前会话中已生成或已上传的图片保存为 PNG、JPEG 或 WebP 文件；需要 Execute 模式，远程项目中文件写入远程主机。'
+    }
   },
   channels: {
     tabs: {