CobbleBrain - AI Dialogue System for Cobblemon
ModCobbleBrain is a mod that gives Pokémon a "brain," allowing them to think, talk, and interact with their surroundings in dynamic ways. Easy to setup and play!
Type
Mod
Modrinth Downloads
3,005
Modrinth ID
YSqiXLE3
Last Updated
Apr 12, 2026
Description
CobbleBrain – AI Dialogue System for Cobblemon
Discord: join the official Cobblemon server, check mods-and-plugins → CobbleBrain.
⚠️Dedicated servers in NEOFORGE are not working in the current versions of the mod... use Syntra Connector and/or other compatibility mods while I resolve this issue.
⚠️Disable the Catch indicator mod (present in Cobbleverse) when playing with cobblebrain.
About
CobbleBrain is a mod that gives Pokémon a “brain,” allowing them to think, talk, and interact dynamically with the world. It integrates artificial intelligence into gameplay, making your companions talk to you, react to battles, protect you from mobs and even cook a delicious steak, all of that using your prompt.
Wild Pokémon can also interact with the player and the world, creating quests, remembering your actions, and reacting to how you treat their species.
Features
- Pokémon can talk to the player and to each other.
- Pokémon can perform actions such as attacking, eating, cooking, repairing tools, growing plants, and more.
- Dialogue influenced by friendship, nature, past interactions, and world conditions.
- Memory system that stores previous interactions.
- Wild Pokémon quests and karma system that react to your actions.
- Raid events triggered when Pokémon species become hostile toward you.
- Fully configurable AI prompts, gameplay settings, and behaviors through the Mod Menu configuration screen.
- Supports cloud AI models (Google AI Studio, OpenAI, OpenRouter, Player2) and local models via LM Studio and Player2.
- Multiplayer compatible, with each player managing their own AI processing.
How to Play
Cloud mode uses external AI providers to process dialogue.
Using Player2
- Install the Player2 app and create an account.
- Run Player2.exe (the app)
Player2 includes a free amount of energy, which can be replenished daily using a spinner. You can also choose the AI model that best fits your needs. In general, more expensive models are smarter. Here's a Tutorial on YouTube!
Using Other Providers
-
Create an account with a provider (examples: Google AI Studio, OpenAI, OpenRouter).
-
Generate an API key from the provider’s dashboard.
-
Choose a model (examples: gemma-3-12b-it, gpt-4.1-mini, anthropic/claude-3.5-sonnet).
-
Open the mod config Menu and fill:
- apiKey: your generated key
- apiBaseUrl: provider’s official URL
- aiModel: ID of the chosen model
Here's a Tutorial on YouTube!
Local mode runs AI models directly on your computer using LM Studio or Player2. WARNING: LOCAL MODELS MAY CAUSE PROBLEMS IF YOU RUN/INSTALL MODELS THAT ARE TOO HEAVY. **Steps:** 1. Install [LM Studio](https://lmstudio.ai) (available for Windows, Mac, Linux). 2. Open LM Studio and set up a folder for storing models. 3. Download a model (examples: LLaMA, Mistral). - 4b–5b models → lightweight, fast, good for simple dialogues. - 7b–8b models → balanced, deeper responses. - 12b+ models → complex, detailed dialogues, requiring significant RAM and GPU. 4. Prefer quantized versions (q4, q5, q8) to reduce resource usage. 5. Start the LM Studio server; it will show a local API address (e.g., `http://localhost:port`). 6. Open the **Config menu** and set: - `apiBaseUrl`: local server address (Example: `http://localhost:1234`) - `aiModel`: ID or name of the model running in the server - `localApiProvider`: `"lmstudio"`- Use the command /mpk <message> to talk to your Pokémon.
- If listenToChat = true, any chat message can be interpreted by the AI.
- Use lowTokenMode for faster and lighter responses.
- Adjust maxShortMemory and maxLongMemory to control how much dialogue memory Pokémon retain.
Usage Recommendations
- Low-end PC / simple laptop → Cloud AI.
- Powerful laptop with dedicated GPU → Cloud AI or lightweight local models (4b–7b).
- Moderate–high-end PC → Cloud AI or robust local models (8b–12b).
- For local models, always prefer quantized versions (q4, q5, q8) to reduce RAM/GPU usage.
Configuration
CobbleBrain can be configured directly in-game through the mod config menu screen. You can open it by pressing Y or typing /cobblebrain openConfig in chat.
There you can customize:
- AI provider settings (apiKey, apiBaseUrl, aiModel)
- AI prompts (instruct and other behavior settings)
- Dialogue behavior
- Wild Pokémon quest chances
- Raid settings
- Pokémon characteristics and personalities
Advanced users can still edit the generated configuration files in the /config folder if needed.
For the full list of variables and explanations, please check the CobbleBrain README on GitHub.
Pokémon Actions
FAQ
Compatibility
Mod Loaders
Game Versions
Screenshots
Similar Mods
Electric Mace
Adds the possibility of enchanting the new Mace with Channeling so that when we hit a mob from above it will be struck b...
Survival Debug Stick
This mod provides players the ability to craft and use Debug Stick in Survival Mode. You also can easily customize its a...
Choconilla [Legacy]
Ideas that I think are unique and special!
Fast-ing
Increasing minimum hunger so you don't have trouble sprinting.
AlmirionCore
AlmirionCore is a lightweight core dependency required for my mods. It does not add gameplay content on its own but prov...
Tinkers' Platinum & Tungsten
Re-adds platinum and tungsten parts to Tinkers' Construct compatibility.