Cascade Wrote the Feature. Now It Needs an Image.
Cascade planned the task, edited seven files, ran the tests, and shipped the feature — all in one session. Then you asked it for a hero image, and the loop broke. Every other step was automated. The image wasn't.
This guide shows how to bring image generation into the same Windsurf session where the code lives — with exact CLI commands and workflow examples for Cascade.
When the Image Gap Blocks Shipping
The gap shows up at specific moments in Cascade workflows:
- Documentation sites: Cascade writes the component structure and copy but the product screenshots are placeholders. The docs ship broken until real images exist.
- Marketing pages: Cascade builds the full page layout but the hero, feature visuals, and social cards are empty slots. The launch waits.
- App assets: Cascade implements the onboarding flow but the illustration assets it references by path don’t exist yet.
- PR descriptions: Cascade opens a clean PR with 40 file changes. The demo image in the description is still missing.
AnyCap provides image generation through a CLI that Windsurf’s terminal can call directly — the same terminal Cascade already uses.
Setup: Adding Image Generation to Windsurf
Step 1: Install AnyCap
curl -fsSL https://anycap.ai/install.sh | sh
export PATH="$HOME/.local/bin:$PATH"
anycap login
AnyCap authenticates once and stores credentials. Every subsequent Windsurf session picks them up automatically — no re-authentication required.
Step 2: Install the Windsurf skill
npx -y skills add anycap-ai/anycap -a windsurf -y
This makes AnyCap's capabilities visible to Cascade's context — so when Windsurf's agent mode is running, it can call AnyCap commands as part of its task loop.
Step 3: Verify
anycap image models
If you see the model list, AnyCap is installed and authenticated. Windsurf can now generate images.
Generating Your First Image in Windsurf
The simplest image generation command:
anycap image generate \
--prompt "your image description" \
--model seedream-5 \
-o output.png
Cascade can run this command directly when you describe the image you need as part of a larger task.
Model Selection for Windsurf Workflows
| Use case | Model | Why |
|---|---|---|
| SaaS UI, dashboard, product screenshot | seedream-5 |
Strong UI rendering, clean dark-mode interfaces |
| Brand hero, marketing asset, announcement | gpt-image-2 |
Versatile composition, strong branded layouts |
| Flux-style precise prompt adherence | flux-kontext-max |
High prompt fidelity, advanced style control |
| Photorealistic product shot | seedream-5 |
High-fidelity product photography style |
| Design-led, abstract, editorial | nano-banana-2 |
Flexible compositional control |
Workflow 1: Build a Feature, Generate Its Documentation Image
What Cascade can now do with AnyCap:
anycap image generate \
--prompt "Clean SaaS analytics dashboard, dark mode, blue accent colors, chart widgets showing user growth metrics, professional product screenshot style" \
--model seedream-5 \
-o docs/images/dashboard-analytics.png
Workflow 2: Generate a Release Announcement Hero
anycap image generate \
--prompt "Software release announcement visual. Dark indigo gradient background. Rounded SaaS cards showing new feature icons. Professional product marketing style. Wide 16:9 format." \
--model gpt-image-2 \
--param aspect_ratio=16:9 \
-o release-hero.png
anycap image generate \
--prompt "Same style as the hero but square format, focused on the central element" \
--model gpt-image-2 \
--param aspect_ratio=1:1 \
-o release-social.png
Workflow 3: Generate Diagrams for Technical Documentation
anycap image generate \
--prompt "Clean technical architecture diagram. Shows three layers: Agent (top), AnyCap CLI (middle), External APIs (bottom). Dark background, light lines connecting components. Minimal labels." \
--model flux-kontext-max \
-o docs/architecture.png
anycap image generate \
--prompt "Simple workflow diagram showing: 1. Agent sends command 2. AnyCap processes 3. API executes 4. Output returned. Linear left-to-right flow. Clean SaaS style, dark background." \
--model seedream-5 \
-o docs/workflow-diagram.png
Windsurf vs Other Coding Agents: Image Generation Comparison
| Agent | Image generation (native) | With AnyCap |
|---|---|---|
| Windsurf / Cascade | ❌ | ✅ |
| Claude Code | ❌ | ✅ |
| Cursor | ❌ | ✅ |
| GitHub Copilot Agent | ❌ | ✅ |
| Codex | ❌ | ✅ |
AnyCap uses the same CLI syntax across all agents.
FAQ
Does image generation work in Windsurf's Cascade flow automatically?
Yes. When AnyCap is installed and authenticated, Windsurf's terminal (and Cascade's agent loop when it has terminal access) can call anycap image generate like any other shell command.
Can I use Windsurf with multiple image models in the same project?
Yes. Each anycap image generate call can specify a different model.
Get Started
# Install AnyCap
curl -fsSL https://anycap.ai/install.sh | sh
export PATH="$HOME/.local/bin:$PATH"
anycap login
# Install for Windsurf
npx -y skills add anycap-ai/anycap -a windsurf -y
# Test: generate your first image
anycap image generate \
--prompt "Clean SaaS product dashboard, dark mode, professional" \
--model seedream-5 \
-o test.png
Windsurf's Cascade now has image generation. Everything Cascade can do with code, it can now do with visuals too.