Quick Start¶
Interactive Chat¶
Start an agentic chat session:
Inside the session, type natural language prompts. The agent can read files, write code, run commands, and search your codebase.
mita> explain the auth module in this project
mita> refactor the database connection to use async
mita> /commit
REPL Commands¶
| Command | Action |
|---|---|
/quit, /exit, /q |
Exit the session |
/clear |
Clear conversation history |
/<skill> |
Invoke a skill (e.g., /commit, /review) |
Single-Shot Prompts¶
For one-off questions without entering an interactive session:
Pipe input from other commands:
Model Management¶
# See what fits your hardware
mita models recommend
# List installed models
mita models list
# Pull a model
mita models pull qwen2.5-coder:14b
# Set as default
mita models default qwen2.5-coder:14b
Next Steps¶
- Memory System — Add persistent context with
MITA.mdfiles - Codebase Indexing — Enable RAG over your codebase
- Skills — Create reusable prompt templates
- Configuration — Customize Mita to your workflow