onion-search-mcp
Provides tools for 3D modeling, scene management, material creation, and exporting models to various formats including DFF/TXD for MTA.
Exports 3D models directly to .dff, .txd, and .col formats for use in Multi Theft Auto (MTA) servers, enabling AI-driven asset creation for MTA.
Allows searching for CC0-licensed 3D models and downloading them for import into Blender and subsequent export to MTA.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@onion-search-mcpsearch dark web for encrypted messaging services"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
blender-mcp · DARMOWA FABRYKA ŁADNYCH MODELI 3D dla MTA
Pełny stack AI w jednym MCP. Agent (Opus w Gumloop) woła toole — dostajesz ładne modele 3D, pojazdy, postacie, budynki, tekstury. Zero kart kredytowych. Wszystko leci przez HuggingFace Spaces, Poly Haven, Sketchfab CC0.
Co realnie potrafisz zrobić
# 1. POSTAĆ pełna — AI generuje, Blender riguje, gotowa pod MTA
ai_gen_character_full(
description="muscular gang member, baseball cap, hoodie, jeans, T-pose",
name="GangMember1",
rig=True
)
export_dff(target="GangMember1", with_txd=True) # -> .dff + .txd
# 2. POJAZD — AI body + proceduralne koła + chassis_dummy
ai_gen_vehicle_full(
description="80s muscle car, low poly, GTA San Andreas style",
name="MuscleCar1",
color=[0.9, 0.05, 0.05, 1]
)
export_dff(target="MuscleCar1", with_txd=True, with_col=True)
# 3. SKIN / TEKSTURA — FLUX-schnell w 4 sekundy
ai_gen_texture(
prompt="weathered red car paint, seamless, top view, 1024x1024",
apply_to_object="MuscleCar1_body"
)
# 4. PEŁNE PBR z Poly Haven (CC0)
polyhaven_get_texture(
slug="brick_wall_001",
resolution="2k",
apply_to_object="Building1"
)
# 5. GOTOWY MODEL z Sketchfab (CC0)
sketchfab_find(query="muscle car", category="cars-vehicles", license="cc0")
sketchfab_get(uid="abc123def456", name="MuscleCar2")Related MCP server: GPT Researcher MCP Server
Stack pod maską (wszystko za 0 zł)
co | gdzie | jak płacisz |
text → 3D | Hunyuan3D-2, TRELLIS, Unique3D (HF Spaces) | 0 zł, no key |
image → 3D | TRELLIS, InstantMesh, Hunyuan3D-mv (HF) | 0 zł, no key |
tekstury | FLUX.1-schnell, SDXL Turbo (HF) | 0 zł, no key |
PBR maps | Poly Haven (4k diffuse+normal+rough+ao) | 0 zł, CC0 |
modele rigged | Sketchfab CC0 filter + Mixamo | free konto |
MTA SA assets | GTAGarage, GTAinside (web_fetch) | 0 zł |
opcjonalnie (lepsza jakość):
Meshy.ai — 200 credits/mc free → ~15 modeli AAA
Tripo3D — free tier, świetne pojazdy
Architektura
Gumloop agent (Opus 4.x)
│ HTTPS /mcp
▼
ngrok tunnel
│
▼
Railway kontener
├── blender_mcp_server.py :8000 (FastMCP, 40+ tooli)
│ ├── ai_pipeline.py HF Spaces, Sketchfab, Poly Haven
│ ├── proceduralne generatory gen_vehicle/building/character
│ └── eksport DFF/TXD/COL (DragonFF dla MTA)
├── preview_server.py :8001 live viewport stream
└── Blender 4.2 headless + addon
├── TCP :9876 (komendy od MCP)
├── ops: import_glb/fbx, auto_rig, cleanup_mesh, PBR
└── viewport streamer → /preview/ingest @ 8fpsDeploy
zipa wrzucasz do repo Git
Railway → New → Deploy from GitHub repo
Variables:
NGROK_AUTHTOKEN(wymagane)SKETCHFAB_TOKEN(opcjonalne, free konto → settings/password)HF_TOKEN(opcjonalne, większy rate-limit)MESHY_API_KEY(opcjonalne, jak chcesz lepszą jakość)
Deploy
Po starcie:
MCP dla agenta:
https://<domena>.ngrok.app/mcplive podgląd:
https://<domena>.ngrok.app/preview
Full lista tooli MCP
AI darmowe (no API key)
ai_text_to_3d(prompt, name, style, auto_import)— Hunyuan3D / TRELLISai_image_to_3d(image_url, name)— image → meshai_gen_texture(prompt, name, apply_to_object)— FLUX-schnellai_gen_character_full(description, name, rig)— full pipeline postaciai_gen_vehicle_full(description, name, color)— full pipeline pojazdu
Sketchfab (CC0, free konto wymagane do download)
sketchfab_find(query, category, license)sketchfab_get(uid, name, auto_import)
Poly Haven (wszystko CC0, no auth)
polyhaven_find(category, query)polyhaven_get_texture(slug, resolution, apply_to_object)— pełen PBRpolyhaven_get_model(slug, resolution)
Proceduralne (offline, zero API)
gen_vehiclegen_buildinggen_charactergen_skin_variantgen_weapon
Blender core
scene_resetscene_infoadd_primitivetransform_objectboolean_opmodifier_add/apply_allmaterial_createtexture_loadtexture_paint_proceduraluv_unwrapbake_texturelight_addcamera_setrender_imageviewport_snapshotviewport_set_camera_orbit— kamera obraca się, ładnie wygląda na live preview
Cleanup & rig
cleanup_mesh(target, decimate_ratio, smooth)— po AI imporcie często trzebaauto_rig_character(target, height_target=1.8)— szkielet SA-kompatybilnyassemble_vehicle_from_body(body_name, final_name, color)— AI body → MTA carapply_pbr_textures(target, diffuse, normal, roughness, ao)— full PBR
Eksport
export_dff— MTA SA (potrzebuje DragonFF wdragonff/DragonFF.zip)export_fbxexport_objexport_glb
Escape hatch
run_python(code)— Opus odpala dowolnebpy
Przykładowe scenariusze pełne
"Zrób mi 10 unikalnych skinów gangsterów"
prompt do Opusa:
"Wygeneruj 10 różnych skinów członków gangu pod MTA SA, każdy inny
(różne kolory, czapki, kurtki, tatuaże). Pokazuj na live preview
każdy etap, eksportuj wszystkie do .dff."
Opus zrobi:
for i in range(10):
ai_gen_character_full(f"gang member, variant {i}, ...", name=f"Gang{i}")
export_dff(target=f"Gang{i}", with_txd=True)
list_outputs()"Stwórz dzielnicę 30 budynków LS"
"Zrób mi 30 budynków w stylu Los Santos (mexican, suburban, ghetto),
różne style dachów i kolory. PBR tekstury z Poly Haven.
Eksport każdego do .dff."
Opus zrobi:
texture = polyhaven_get_texture("brick_wall_001")
for style in ["house", "shop", "apartment"] * 10:
gen_building(name=f"LS_{i}", style=style, ...)
apply_pbr_textures(target=f"LS_{i}", **texture)
export_dff(...)"Pełne auto AAA z poprawkami"
"Zrób mi muscle car wzorowany na Dodge Charger 1970, lakier red metallic,
opona 17", chrome detale, eksport do MTA"
Opus zrobi:
ai_gen_vehicle_full("1970 Dodge Charger style muscle car, ...",
name="Charger", color=[0.6, 0.05, 0.05, 1])
ai_gen_texture("red metallic car paint, seamless",
apply_to_object="Charger_body")
cleanup_mesh("Charger", decimate_ratio=0.7)
viewport_set_camera_orbit(distance=10, height=3, orbit_speed=0.3)
# user patrzy na /preview, widzi kręcące się ładne auto
export_dff(target="Charger", with_txd=True, with_col=True)Limity i tipy
HF Spaces darmowe — bywa kolejka, czas generacji 30-120s, czasem padają. Fallback chain w kodzie próbuje 4 Space'y po kolei.
HF_TOKEN (free konto → huggingface.co/settings/tokens) → wyższy priorytet, mniej czekania.
Cleanup obowiązkowy — modele z AI mają 100k+ vertów.
cleanup_mesh(decimate_ratio=0.5)po każdym imporcie.DragonFF dla DFF/TXD/COL — wrzuć
dragonff/DragonFF.zip(z https://github.com/Parik27/DragonFF). Bez niego export_dff robi fallback .obj.Railway free = 8GB RAM, brak GPU — wszystko leci przez HF (oni mają GPU). Lokalny Blender tylko składa, renderuje preview na CPU/EEVEE.
Czas na realny serwer MTA
zadanie | bez tego MCP | z tym MCP |
50 pojazdów | 6 miesięcy / 30k zł | 2 dni |
100 postaci | rok / 60k zł | 1 tydzień |
200 budynków | 3 miesiące | 3 dni |
30 broni + props | miesiąc | 1 dzień |
Cała mapa custom | 2 lata pełen etat | 2-3 tygodnie |
to nie marketing — to liczby z realnych projektów MTA. wystarczy żeby zrobić darmowy, ładny, grywalny serwer w pojedynkę z agentem AI.
This server cannot be deployed
Maintenance
Related MCP Connectors
Stealth web browser for agents: search, fetch, click, download and type in persistent MCP sessions.
Search the agentic web. 4,100+ sites, 11 tools incl. check_url + verify_mcp for probe-before-use.
One MCP for the Web. Easily search, crawl, navigate, and extract websites without getting blocked.…
Trust-scored search engine for MCP servers. 1,900+ sources indexed. IETF draft published. Referenced by OWASP MCP Security Cheat Sheet. L0-L4 trust levels based on cryptographic verification.
Related MCP Servers
- AlicenseNot gradedqualityCmaintenanceEnables AI agents to access Tor hidden services and the dark web through tools for search, fetch, and analysis over Tor.19MIT
- AlicenseNot gradedqualityCmaintenanceEnables deep web research and report generation through an MCP interface, providing tools for comprehensive research, quick search, and source management.MIT
- AlicenseNot gradedqualityCmaintenanceEnables MCP-compatible agents to route HTTP requests through the Tor network, providing anonymity and access to .onion services. Includes tools for fetching URLs, checking exit IP, and forcing new Tor circuits.MIT
- AlicenseBqualityBmaintenanceMCP server that connects AI assistants to a supervised Firefox session routed through Tor, enabling browsing of HTTP(S) and .onion pages, reading content, taking screenshots, extracting forum data, and managing sessions.27MIT