01
Vision Chat
Work naturally with text and images while keeping conversation history on your computer.
One local workspace
More useful than another empty chat box.
Basalt organizes local AI into durable workflows: conversations you can find again and focused Skills the model can load only when needed.
02
Custom Skills
Define reusable expertise with a name, description, and detailed local instructions.
03
Curated Models
Compare focused model options with clear size, quality, memory, and disk requirements.
04
Device Checks
Basalt blocks downloads when your computer lacks the memory needed to run the model.
05
On-demand Context
Skill details enter the conversation only after the local model requests them by name.
06
Generation Control
Adjust the system prompt, context, creativity, token limits, and performance settings.
The real Windows application
See the complete local AI workspace.
These are current Basalt screens, shown at their natural 16:9 proportion so the interface remains clear on every screen size.
A deliberate three-step flow
Download carefully. Work locally. Reuse expertise.
01
Check the device
See the complete memory and disk requirement before a large model download can begin.
02
Build your workspace
Use private chats and reusable Skills instead of recreating the same instructions.
03
Shape the assistant
Edit the default system prompt and let the local model fetch relevant Skill details on demand.
Practical local AI guides
Solve the problem, then choose the tool.
Detailed tutorials about private AI, GGUF model selection, multi-image analysis, and dependable AI workflows. Each guide is useful even if you use another local AI application.
Privacy guide · 12 min
READ THE GUIDE →
How to Run a Private AI Assistant on Windows
A practical path from hardware checks to a local, testable workflow for sensitive documents and images.
Troubleshooting · 11 min
READ THE GUIDE →
Why Local AI Runs Out of Memory—and How to Fix It
Understand weights, context, image projectors, and GGUF quantization before you download the wrong model.
Vision tutorial · 10 min
READ THE GUIDE →
How to Analyze Multiple Images with a Local AI Model
Prepare comparable images, write an evidence-first prompt, and verify what the vision model actually saw.
Workflow tutorial · 13 min
READ THE GUIDE →
How to Build Reusable AI Workflows That Stay Reliable
Separate permanent rules, reusable expertise, task context, and tool permissions so assistants behave consistently.
Straight answers
Frequently asked questions.
Does Basalt require an account?
No. Basalt does not provide an account system, cloud history, or cloud synchronization.
Where are my chats and Skills stored?
Your conversations, Skills, settings, and downloaded models are stored locally on your device.
Is Basalt completely offline?
Core inference runs locally. An internet connection is needed only for user-directed actions such as downloading a model or opening an external support or Store link.
How does Basalt choose a model for my PC?
You choose the model. Before download, Basalt compares its complete memory requirement with your system memory and checks available disk space.
How do Skills affect context?
The model always sees each Skill name and description. Full Skill instructions enter the conversation only when the local model calls get_skill with that name.
Can I change the assistant instructions?
Yes. You can edit the system prompt in Settings or restore the Basalt default at any time.
Which languages are available?
The interface supports English, Simplified Chinese, Traditional Chinese, Japanese, Korean, French, German, Spanish, and Italian.
How can I get support?
Email support@iamdev.me with questions, issues, or feedback.
Make local AI part of your real workflow.
Private chat, reusable tools, and models chosen for your device.