forge
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| MESHY_API_KEY | No | Optional API key for Meshy image and 3D models. | |
| GEMINI_API_KEY | No | Optional API key for Gemini image models. | |
| OPENAI_API_KEY | No | Optional API key for OpenAI chat models. | |
| ANTHROPIC_API_KEY | No | Optional API key for Anthropic chat models. | |
| ELEVENLABS_API_KEY | No | Optional API key for ElevenLabs image models, including GPT Image, Seedream, and Nano Banana. |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| list_assetsA | List all assets and the lineage DAG in a Forge project, optionally filtered by type, status, approval_state, or tag. |
| get_assetB | Get full metadata, prompt lineage, QC report, Critic report, parents, and children for a single asset. |
| run_verbC | Run a Forge verb (variant, make_3d, turnaround, extract_part, kitbash, retexture, icon_maker, sprite_sheet, tileable_texture, batch_from_list, qc_check) on parent asset(s). |
| run_recipeC | Run any declarative YAML/JSON recipe by ID with loops, critic checks, and budget guards. |
| create_recipeC | Register a new declarative YAML or JSON recipe without touching code. |
| list_style_packsB | List all available built-in and auto-scanned Style Packs. |
| get_project_bibleB | Get the Project Bible (art style, palette, polycount budget, target engine, autonomy mode, spend caps). |
| update_project_bibleC | Update fields in the Project Bible. |
| rate_assetC | Rate or swipe an asset (favorite, rejected, keep, more_like_this) and record taste signals. |
| queue_statusC | Inspect live job queue status, progress, and current budget spend. |
| get_budgetB | Get current project spend, per-run cap, project cap, and remaining USD budget. |
| exportC | Export engine-ready folders and metadata for Godot, Unity, or Unreal (never includes API keys). |
| scan_game_projectC | Read-only scan of an existing game project folder (Godot 4 primary) to build a Game Profile with evidence & confidence. |
| get_game_profileB | Retrieve the scanned Game Profile (including evidence, confidence levels, catalog, and wishlist). |
| update_game_profileC | Update and approve fields in the Game Profile and lock them into the Project Bible. |
| list_asset_wishlistC | List prioritized missing assets, placeholders, and TODO art items discovered by Game Scan. |
| find_placeholdersC | List placeholder or temp assets found in the scanned game project for replacement. |
| propose_assets_forC | Synthesize context-aware prompts (icon, 3D prop, variants) for a specific catalog item ID. |
| writeback_previewC | Stage assets for export into the game project and return a non-destructive diff preview. |
| writeback_applyC | Write staged assets and engine glue stubs into the game project (never overwrites existing files). |
| run_agent_briefB | Execute a high-level natural language agent brief end to end within budget (e.g. 'Make 20 fantasy potions matching the project bible, then 3D for the 5 best'). |
| undo_actionC | Undo the last user or agent action (or a specific activity_id). |
| read_workspace_docC | Read a shared Markdown file in the project workspace (e.g. GAME.md, ASSET_REQUIREMENTS.md, HANDOFFS.md, KNOWLEDGE.md, DELIVERY_MANIFEST.json). |
| write_workspace_docB | Write or update a shared Markdown file in the project workspace (e.g. GAME.md or ASSET_REQUIREMENTS.md). Editing GAME.md automatically syncs the Project Bible. |
| list_asset_requirementsB | Parse workspace/ASSET_REQUIREMENTS.md and return open [ ], in-progress [~], and completed [x] asset requirements with SHA-256 receipts. |
| add_asset_requirementB | Add a new asset requirement row ( |
| fulfill_asset_requirementsC | Automatically generate open [ ] asset requirements in workspace/ASSET_REQUIREMENTS.md, compute SHA-256 delivery receipts, mark them [x], and append a handoff note to HANDOFFS.md. |
| create_blueprint_folderA | Create a nested blueprint folder under a category, such as characters/bosses/infernal. Does not create or delete blueprints. |
| delete_blueprintA | Permanently delete one blueprint. Requires confirm=true. Generated assets stay in the vault unless delete_generated_assets is also true. |
| post_agent_handoffB | Append a structured handoff message to workspace/HANDOFFS.md to communicate between the external game-coding agent and Forge. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 30 tools
Most tools target clearly distinct resources or workflow stages, and descriptions distinguish similar pairs such as writeback_preview/writeback_apply and read_workspace_doc/get_project_bible. Some ambiguity remains between high-level execution tools like run_verb, run_recipe, and run_agent_brief, where an agent must infer the right abstraction level.
The set mostly follows a predictable snake_case verb_noun convention: get_asset, list_assets, update_project_bible, writeback_apply, etc. Minor deviations such as queue_status and the bare export keep it from being perfectly consistent, but overall it is highly readable.
At 30 tools, the surface is well beyond a focused toolset and covers many subdomains: generation, scanning, requirements, writeback, workspace docs, budgeting, recipes, and blueprints. Although the domain is broad, this many tools creates selection overhead and increases the chance of choosing the wrong workflow entry point.
Core asset-generation, project-bible, game-profile, requirement, budgeting, and writeback workflows are covered. However, lifecycle gaps are notable: blueprints have create-folder and delete but no list/get/update, recipes can be created and run but not listed or deleted, and assets have no update/delete operations.