diff --git a/src/components/studio/views/ai-harness-view-coverage.test.tsx b/src/components/studio/views/ai-harness-view-coverage.test.tsx
new file mode 100644
index 00000000..02c36476
--- /dev/null
+++ b/src/components/studio/views/ai-harness-view-coverage.test.tsx
@@ -0,0 +1,340 @@
+import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest'
+import { act } from 'react'
+import type { ReactNode } from 'react'
+import { render, screen, fireEvent, cleanup, waitFor } from '@testing-library/react'
+
+// Hoisted so vi.mock factories (hoisted above imports) can reference these mocks.
+const aiMocks = vi.hoisted(() => {
+  const mockSendChatStream = vi.fn()
+  const mockFetchOllamaModels = vi.fn(() => Promise.resolve(['llama3']))
+  const mockBuildMessages = vi.fn(() => [])
+  const mockLoadAISettings = vi.fn(() =>
+    Promise.resolve({
+      provider: 'openrouter',
+      model: 'openrouter/free',
+      apiKey: '',
+      augmentWithLocal: true,
+      ollamaCpuOnly: false,
+      allowWebResearch: false,
+      ollamaBaseUrl: 'http://localhost:11434',
+    }),
+  )
+  const mockSaveAISettings = vi.fn((_s: Record<string, unknown>) => Promise.resolve())
+  const mockFetchUrls = vi.fn(() =>
+    Promise.resolve([{ success: true, url: 'https://example.com' }]),
+  )
+  return {
+    mockSendChatStream,
+    mockFetchOllamaModels,
+    mockBuildMessages,
+    mockLoadAISettings,
+    mockSaveAISettings,
+    mockFetchUrls,
+  }
+})
+
+const rateLimitState = vi.hoisted(() => ({
+  decision: { allowed: true, count: 0, limit: 10, retryAfterMs: 5000 },
+}))
+
+const storeState = vi.hoisted(() => ({
+  mockEntities: [
+    {
+      id: 'ent-1',
+      name: 'Test Entity',
+      type: 'concept' as const,
+      description: 'A test concept',
+      content: '# Hello',
+      tags: [],
+      createdAt: new Date().toISOString(),
+      updatedAt: new Date().toISOString(),
+      links: [],
+    },
+  ],
+}))
+
+vi.mock('framer-motion', () => ({
+  motion: {
+    div: ({
+      children,
+      initial: _i,
+      animate: _a,
+      transition: _t,
+      ...props
+    }: { children?: ReactNode; [key: string]: unknown }) => (
+      <div {...(props as React.HTMLAttributes<HTMLDivElement>)}>{children}</div>
+    ),
+  },
+  AnimatePresence: ({ children }: { children?: ReactNode }) => <>{children}</>,
+}))
+
+vi.mock('lucide-react', () => {
+  const Icon = ({ className }: { className?: string }) => (
+    <span data-testid="icon" className={className} />
+  )
+  return {
+    FlaskConical: Icon,
+    Bot: Icon,
+    User: Icon,
+    Send: Icon,
+    Settings: Icon,
+    Database: Icon,
+    Key: Icon,
+    Cpu: Icon,
+    Plug: Icon,
+    Check: Icon,
+    BookOpen: Icon,
+    Sparkles: Icon,
+    Zap: Icon,
+    RefreshCw: Icon,
+    Globe: Icon,
+  }
+})
+
+vi.mock('sonner', () => ({
+  toast: { success: vi.fn(), error: vi.fn(), info: vi.fn(), warning: vi.fn() },
+}))
+
+vi.mock('@/lib/utils', () => ({
+  cn: (...args: (string | undefined | false | null)[]) => args.filter(Boolean).join(' '),
+}))
+
+vi.mock('@/lib/studio/ai-settings', () => ({
+  loadAISettings: () => aiMocks.mockLoadAISettings(),
+  saveAISettings: (s: Record<string, unknown>) => aiMocks.mockSaveAISettings(s),
+}))
+
+vi.mock('@/lib/studio/use-reduced-motion', () => ({
+  useReducedMotion: () => false,
+}))
+
+vi.mock('@/lib/ai', () => ({
+  sendChatStream: aiMocks.mockSendChatStream,
+  fetchOllamaModels: aiMocks.mockFetchOllamaModels,
+  buildMessages: aiMocks.mockBuildMessages,
+  useRateLimiter: () => ({ canRequest: () => rateLimitState.decision }),
+  OPENROUTER_ROUTERS: [{ slug: 'openrouter/auto', display_name: 'Auto Router' }],
+  OPENROUTER_MODELS: [{ slug: 'openai/gpt-4o-mini', display_name: 'GPT-4o Mini' }],
+  OPENROUTER_DEFAULT_TARGETS: [
+    { slug: 'openrouter/auto', display_name: 'Auto Router' },
+    { slug: 'openai/gpt-4o-mini', display_name: 'GPT-4o Mini' },
+  ],
+}))
+
+vi.mock('@/lib/ai/types', () => ({
+  DEFAULT_MODEL: { openrouter: 'openrouter/free', ollama: 'llama3' },
+  OLLAMA_DEFAULT_MODELS: ['llama3', 'mistral'],
+  DEFAULT_OLLAMA_BASE_URL: 'http://localhost:11434',
+}))
+
+vi.mock('@/lib/ai/research', () => ({
+  extractUrls: (text: string) => (text.includes('http') ? ['https://example.com'] : []),
+  fetchUrls: aiMocks.mockFetchUrls,
+}))
+
+vi.mock('./ai-harness-settings', () => ({
+  PROVIDERS: [
+    { id: 'openrouter', label: 'OpenRouter', models: ['openrouter/free'], requiresKey: true },
+    { id: 'ollama', label: 'Ollama (local)', models: ['llama3'], requiresKey: false },
+  ],
+  Field: ({ label, children }: { label: string; children?: ReactNode }) => (
+    <div data-testid="field">
+      <span>{label}</span>
+      {children}
+    </div>
+  ),
+}))
+
+vi.mock('../ui/shared-primitives', () => ({
+  SwitchToggle: ({
+    label,
+    checked,
+    onToggle,
+  }: {
+    label: string
+    checked?: boolean
+    onToggle?: () => void
+  }) => (
+    <button
+      data-testid="switch-toggle"
+      data-checked={String(Boolean(checked))}
+      onClick={() => onToggle?.()}
+    >
+      {label}
+    </button>
+  ),
+}))
+
+vi.mock('@/lib/studio/store', () => ({
+  useStudioStore: (selector: (s: Record<string, unknown>) => unknown) =>
+    selector({
+      entities: storeState.mockEntities,
+      claims: [],
+    }),
+}))
+
+import { AIHarnessView } from './ai-harness-view'
+
+const sendMessage = async (text: string) => {
+  const textarea = screen.getByPlaceholderText(/Ask the AI agent/)
+  fireEvent.change(textarea, { target: { value: text } })
+  fireEvent.click(screen.getByLabelText('Send'))
+  await act(async () => {
+    await Promise.resolve()
+  })
+  await act(async () => {
+    await Promise.resolve()
+  })
+}
+
+describe('AIHarnessView branch coverage', () => {
+  beforeEach(() => {
+    vi.clearAllMocks()
+    rateLimitState.decision = { allowed: true, count: 0, limit: 10, retryAfterMs: 5000 }
+    aiMocks.mockSendChatStream.mockImplementation(async () => {})
+    aiMocks.mockFetchOllamaModels.mockImplementation(() => Promise.resolve(['llama3']))
+    aiMocks.mockBuildMessages.mockReturnValue([])
+  })
+
+  afterEach(() => {
+    cleanup()
+    vi.restoreAllMocks()
+  })
+
+  it('does not send when input is empty', async () => {
+    await act(async () => {
+      render(<AIHarnessView />)
+    })
+    fireEvent.click(screen.getByLabelText('Send'))
+    expect(aiMocks.mockSendChatStream).not.toHaveBeenCalled()
+  })
+
+  it('shows error toast when no API key and provider requires it', async () => {
+    await act(async () => {
+      render(<AIHarnessView />)
+    })
+    await sendMessage('Hello')
+    const { toast } = await import('sonner')
+    expect(toast.error).toHaveBeenCalledWith('Set an API key in settings to send messages.')
+    expect(aiMocks.mockSendChatStream).not.toHaveBeenCalled()
+  })
+
+  it('sends message and streams response', async () => {
+    aiMocks.mockSendChatStream.mockImplementation(
+      async (_opts: unknown, onChunk: (c: string) => void) => {
+        onChunk('chunk1')
+      },
+    )
+    await act(async () => {
+      render(<AIHarnessView />)
+    })
+    // Open settings, set API key, then send
+    fireEvent.click(screen.getByText('Show settings'))
+    const keyInput = screen.getByPlaceholderText('sk-or-\u2026')
+    fireEvent.change(keyInput, { target: { value: 'test-key' } })
+    await sendMessage('Hello')
+    await waitFor(() => {
+      expect(aiMocks.mockSendChatStream).toHaveBeenCalled()
+    })
+    expect(aiMocks.mockBuildMessages).toHaveBeenCalled()
+  })
+
+  it('adds error message when sendChatStream throws', async () => {
+    aiMocks.mockSendChatStream.mockRejectedValue(new Error('boom'))
+    await act(async () => {
+      render(<AIHarnessView />)
+    })
+    fireEvent.click(screen.getByText('Show settings'))
+    const keyInput = screen.getByPlaceholderText('sk-or-\u2026')
+    fireEvent.change(keyInput, { target: { value: 'test-key' } })
+    await sendMessage('Hello')
+    await waitFor(() => {
+      expect(screen.getByText(/\[Error\] boom/)).toBeDefined()
+    })
+  })
+
+  it('adds rate-limited assistant message when rate limited', async () => {
+    rateLimitState.decision = { allowed: false, count: 10, limit: 10, retryAfterMs: 5000 }
+    await act(async () => {
+      render(<AIHarnessView />)
+    })
+    fireEvent.click(screen.getByText('Show settings'))
+    const keyInput = screen.getByPlaceholderText('sk-or-\u2026')
+    fireEvent.change(keyInput, { target: { value: 'test-key' } })
+    await sendMessage('Hello')
+    await waitFor(() => {
+      expect(screen.getByText(/rate-limited/)).toBeDefined()
+    })
+  })
+
+  it('fetches URLs when web research is allowed', async () => {
+    aiMocks.mockSendChatStream.mockImplementation(async () => {})
+    await act(async () => {
+      render(<AIHarnessView />)
+    })
+    fireEvent.click(screen.getByText('Show settings'))
+    const keyInput = screen.getByPlaceholderText('sk-or-\u2026')
+    fireEvent.change(keyInput, { target: { value: 'test-key' } })
+    // Click the "Allow web research" toggle explicitly
+    fireEvent.click(screen.getByText('Allow web research'))
+    await sendMessage('Check https://example.com')
+    await waitFor(() => {
+      expect(aiMocks.mockSendChatStream).toHaveBeenCalled()
+    })
+    expect(aiMocks.mockFetchUrls).toHaveBeenCalled()
+  })
+
+  it('handles abort error without adding message', async () => {
+    const abortError = new DOMException('Aborted', 'AbortError')
+    aiMocks.mockSendChatStream.mockRejectedValue(abortError)
+    await act(async () => {
+      render(<AIHarnessView />)
+    })
+    fireEvent.click(screen.getByText('Show settings'))
+    const keyInput = screen.getByPlaceholderText('sk-or-\u2026')
+    fireEvent.change(keyInput, { target: { value: 'test-key' } })
+    await sendMessage('Hello')
+    // Wait for loading to finish (abort path returns early without an error message)
+    await waitFor(() => {
+      expect(screen.getByText('Ready')).toBeDefined()
+    })
+    expect(screen.queryByText(/\[Error\]/)).toBeNull()
+  })
+
+  it('saves settings after load', async () => {
+    await act(async () => {
+      render(<AIHarnessView />)
+    })
+    await waitFor(() => {
+      expect(aiMocks.mockSaveAISettings).toHaveBeenCalled()
+    })
+  })
+
+  it('fetches ollama models on refresh button click', async () => {
+    await act(async () => {
+      render(<AIHarnessView />)
+    })
+    fireEvent.click(screen.getByText('Show settings'))
+    // Switch provider to ollama so the refresh button appears
+    const providerSelect = screen.getAllByRole('combobox')[0]
+    fireEvent.change(providerSelect, { target: { value: 'ollama' } })
+    fireEvent.click(screen.getByLabelText('Refresh Ollama models'))
+    await waitFor(() => {
+      expect(aiMocks.mockFetchOllamaModels).toHaveBeenCalled()
+    })
+  })
+
+  it('falls back to default models when fetch returns none', async () => {
+    aiMocks.mockFetchOllamaModels.mockResolvedValue([])
+    await act(async () => {
+      render(<AIHarnessView />)
+    })
+    fireEvent.click(screen.getByText('Show settings'))
+    const providerSelect = screen.getAllByRole('combobox')[0]
+    fireEvent.change(providerSelect, { target: { value: 'ollama' } })
+    fireEvent.click(screen.getByLabelText('Refresh Ollama models'))
+    await waitFor(() => {
+      expect(aiMocks.mockFetchOllamaModels).toHaveBeenCalled()
+    })
+  })
+})