Game Builder Workflow

This guide walks through a complete game design session using the game-design-session MCP prompt. The prompt guides the AI assistant to ask you one question at a time, building up the design progressively.

Prerequisites

Before starting, make sure:

• You have a valid API key
• The MCP server is configured in your AI assistant
• You have an idea for a game

Starting a Session

In your AI assistant, start with:

"Let's design a game using the game builder."

Or use the game-design-session prompt directly.

The MCP will respond with a structured session walkthrough covering all 5 stages.

---

Step-by-Step Walkthrough

Below is a complete example of a game design session. This shows the conversation flow between you and the AI assistant, with the MCP tools being called in the background.

Stage 1: Vision

AI: "Let's start with your game. What's the working title?"

You: "Echoes of the Wild"

AI calls create-game-design(title: "Echoes of the Wild")

AI: "Describe your dream game in one sentence."

You: "A creature-collector metroidvania where you capture and evolve monsters to unlock new abilities and explore a vast interconnected world."

AI calls set-game-vision(game_design_id, elevator_pitch, target_platforms: ["pc"], target_playtime_minutes: 600)

AI: "What genre or genre-mix?"

You: "Creature collector + metroidvania"

AI calls set-game-genre(game_design_id, genre: ["creature-collector", "metroidvania"])

AI: "What games inspire this project, and which aspects do you want to learn from?"

You: "Hollow Knight (art, atmosphere, level design), Pokemon (creature types, evolution), Ori (movement, visual polish)"

AI calls set-game-comparables(game_design_id, comparables: [{name, aspects}...])

AI: "Tell me the story and core theme."

You: "In a world where ancient monsters sealed a great evil, a young tamer must collect all the legendary creatures to prevent the seal from breaking. Theme: unity and balance."

AI calls set-game-story(game_design_id, story, theme: "unity")

AI suggests: "Now let's extract the visual style from your references."

AI calls extract-style-dna(game_design_id) then generate-art-direction-bible(game_design_id)

AI: "Your style DNA is extracted. The art direction bible suggests: side-view perspective, high contrast, melancholic mood. I'm suggesting a moody purple-blue color script with dark forest tones. Sound good?"

You: "Perfect."

Stage 2: Scope

AI: "Let's set scope constraints. Team size?"

You: "Solo dev"

AI: "Development timeframe?"

You: "One year"

AI: "Quality target?"

You: "Indie quality — polished but not AAA"

AI calls set-game-scope(game_design_id, team_size: "solo", timeframe: "year", quality: "indie") AI calls estimate-game-scope(game_design_id)

AI: "Based on your constraints, here's the estimate:

• Max enemy types: 8
• Max biomes: 3
• Max NPCs: 15
• Estimated total assets: ~340

This is realistic for a solo dev in one year. I recommend running scope compression to fine-tune."

AI calls compress-game-scope(game_design_id)

AI: "Your ideal scope would need a team of 6 over 3 years. I've compressed it to fit solo/year. The main changes: enemy types from 30 to 8, NPCs from 40 to 15. Still happy?"

You: "Yes, that works."

Stage 3: Design Bible

AI: "Let's build the core loop."

AI calls generate-core-loop(game_design_id)

AI: "Core loop generated: Explore biomes → Encounter wild creatures → Battle and capture → Evolve and unlock abilities → Access new areas. Secondary loops: Collection (find+evolve all creatures), Metroidvania backtracking (new abilities unlock old paths).

Ready to lock creative direction?"

AI calls lock-creative-direction(game_design_id, visual_style: "pixel-art", tone: "adventurous", music_style: "orchestral", narrative_style: "lore-heavy", confirm: true)

AI: "Direction locked. Now generating the full game bible."

AI calls generate-game-bible(game_design_id)

AI: "Game bible complete with progression, economy, combat, narrative, controls, and UX sections."

Stage 4: Asset Planning

AI: "Ready to discover all assets your game needs?"

AI calls discover-game-assets(game_design_id)

AI: "I found 478 assets across all categories. Breakdown:

Feature-triggered assets from your story (no extra mechanics detected, but creature evolution added evolution animations and UI).

Generating dependency graph..."

AI calls generate-asset-dependency-graph(game_design_id)

AI: "Dependency graph built: 183 nodes with 247 edges. Key chains:

• Hero → idle/walk/jump/climb animations → creature_ball sprite
• Each creature → battle sprite → evolution animation → cry SFX
• Each biome → tileset → background → ambient SFX"

AI calls generate-asset-prompts(game_design_id)

AI: "Asset prompts generated for all 478 assets. Each prompt is pre-filled with your style DNA and genre context."

AI calls estimate-asset-budget(game_design_id)

AI: "Asset budget: 478 total assets. Medium risk: large creature roster (35 creatures) increases animation workload. Consider starting with 20 creatures for the vertical slice."

Stage 5: Review & Export

AI calls review-game-design(game_design_id)

AI: "Design Review Results

Ship probability: 68%

Issues: High scope risk — consider reducing creature count or cutting secondary biomes.

Recommendations:

1. Start with 20 creatures instead of 35
2. Cut 1 biome for the initial release
3. Consider procedural creature generation"

AI calls check-design-health(game_design_id)

AI: "All design health checks passed! Scope, fun, lore, and art are aligned."

AI calls analyze-competitive-positioning(game_design_id)

AI: "Market Analysis

• Steam tags: Creature Collector, Metroidvania, Pixel Graphics, Adventure, Exploration
• Market gap: Very few games combine creature collection with metroidvania exploration
• Direct competitors: Cassette Beasts, Temtem, Monster Sanctuary
• Recommendation: Lead marketing with the unique genre combination"

AI calls export-game-repository(game_design_id)

AI: "Design repository exported! 24 files generated covering the complete game design."

Next Steps

After completing the game design session, you can:

1. Generate assets: Use the standard tools (generate-image, create-character, create-tileset, etc.) with the prompts from your asset manifest
2. Feed coding agents: Use generate-coding-context to produce spec files for Cursor/Claude Code
3. Plan development: Use plan-vertical-slice to start with the minimum playable demo
4. Iterate: Call set-game-* tools to update any section, then re-run review and export
5. Share: The exported repository is a complete game design document you can share with a team

Starting from Scratch

If you don't have a detailed concept yet, just say:

"Let's design a game, but I don't know what I want yet."

The AI assistant will help you brainstorm by asking about your favorite games, preferred genres, and what kind of experience you want to create. It can suggest genre combinations, mechanics, and themes based on your answers.