Mia Agent Mathew Manoj

Use this command to install Mia Agent with WinGet:

winget install --id=Mach50.MiaAgent -e

Mia Agent is a personal AI assistant designed to serve as an unobtrusive and configurable tool for users seeking a coding and productivity companion. Built with principles of user agency and minimalism, Mia grows alongside its users, offering support without replacing existing workflows.

Key Features:

Minimal Terminal UI: A clean interface that prioritizes simplicity and ease of use.
Multi-turn Agent Loop: Enables ongoing conversations and task continuity.
Agent Memory & History: Maintains context and records interactions for better assistance over time.
Markdown Rendering: Supports rich content display directly in the terminal.
OpenAI Compatible API Access: Integrates with OpenRouter for language model functionality.
Web Search & Extraction: Leverages Tavily for real-time information gathering.
Toolset Integration: Includes capabilities like file system interaction, Python execution, and shell commands.

Audience & Benefit: Ideal for developers, researchers, and individuals seeking to enhance their workflows with AI-driven assistance. Mia provides a customizable platform that streamlines tasks, supports decision-making, and integrates seamlessly into existing tools and processes. Its design emphasizes user control and adaptability, making it a versatile addition to any productivity setup.

Mia Agent can be installed via winget on Windows, ensuring quick and straightforward setup for users.

Mia Agent

Mia is a coding and personal assistant AI agent designed to be unobtrusive, configurable and easy to use. Mia grows with you.

This project was inspired by Hermes Agent. Although it borrows ideas from it, this is not a direct derivation or modification of the project.

Principles

The agent should never write to the system without getting permission from the user.
Guardrails should be implemented in code, not in prompts.
The agent should act as an additional tool to the user not as a replacement for everything they already use.

Features

Minimal terminal UI powered by ratatui
Multi turn agent loop with configurable iteration limit
Agent memory (persistent across sessions)
Session history with auto-loading of the last session (mia session list / mia session clear)
Full markdown rendering in terminal (thanks to termimad)
Streaming responses (enabled by default, with token usage display)
OpenAI Compatible API access with support for multiple providers:
- OpenRouter (default, recommended)
- OpenCode Zen
- Google AI Studio
- Groq
- Cerebras
- Local LLMs (Ollama, LMStudio, llama.cpp, etc.)
Web search and extract using Tavily
Document conversion using pandoc (doc_convert tool)
Document creation from markdown/latex (doc_create tool)
Legacy reedline-based TUI available via the hidden mia old-tui command

Tools

Currently Mia has the following tools:

Tool	Description	Gated
📅	`datetime`	Get the current date and time
📁	`fs_list_dir`	List files in a directory
📖	`fs_read_file`	Read a file from the filesystem
🔍	`fs_grep_files`	Search file contents
🧭	`fs_search_dirs`	Search for files in a directory
✍️	`fs_write_file`	Write content to a file
🐍	`exec_python`	Execute Python 3 code
🐚	`exec_shell`	Run shell commands
🧠	`memory`	Manage memory about the user and the agent
🪏	`web_extract`	Extract content from a URL
🌐	`web_search`	Search the web
📄	`doc_create`	Create pdf, docx and other documents
📃	`doc_convert`	Convert any document to another using pandoc

Tools that are gated require explicit user permission before executing.

Installation

Option A: Prebuilt binaries

Prebuilt binaries for Linux (x86_64) and Windows (x86_64) are published as GitHub releases from the releases page.

Open a new terminal and run

mia setup

to configure your agent and start using it.

Option B: Using cargo (Windows/Linux)

Mia requires rust nightly

rustup toolchain install nightly
rustup default nightly

Cargo can fetch, build and install mia.

cargo +nightly install --git https://github.com/mastermach50/mia-agent

Run

mia setup

to configure your agent and start using it.

Option C: Using nix flakes (NixOS/Linux)

Run the given command to test out mia on your system.

nix run github:mastermach50/mia-agent

To install it, add to your system packages using the same flake.

Configuration

Mia supports multiple LLM providers:

OpenRouter (default, recommended) - can be obtained for free, there are also free models
OpenCode Zen - competitive pricing, free credits
Google AI Studio - free tier available
Groq - fast inference API
Cerebras - large model inference
Local LLMs - Ollama, LMStudio, llama.cpp, and other OpenAI-compatible local servers

Run mia setup to configure your agent interactively. mia setup supports section selection: --model, --tui and --agent to run only that part of the setup.
For advanced configuration, you can manually edit the ~/.mia/.env file on Linux or C:\Users\\.mia\.env on Windows to add your API key:

OPENROUTER_API_KEY=

Or for other providers, set the appropriate environment variable.

(Optional) Add your Tavily API key (can be obtained for free and has a reasonable number of free searches per month) to .mia/.env to enable the web_search and web_extract tools.

TAVILY_API_KEY=

(Optional) Run mia tools to see if all the tools are usable and find which components are missing if any.

Currently the external tools required by mia are

Usage

CLI

Use mia --help to access the full cli help menu.

TUI

The TUI supports certain commands. Use /help while in the TUI to see them.

Available Commands

mia tui - Start the main terminal UI (ratatui-based)
mia old-tui - Start the legacy reedline-based TUI (hidden command)
mia setup - Interactive setup wizard
mia tools - Check tool availability
mia model list - List available models
mia model list --free - List only free models (price filter)
mia model list --min-context - Filter by minimum context length (supports human readable values like 128k, 4M)
mia model list --max-price - Filter by maximum price per million tokens
mia model show - Show current model info
mia sessions list - List all sessions
mia session clear - Clear all sessions (prompts for confirmation)

Development

MSRV

Mia uses rust nightly because the code relies on unstable Rust features (iter_intersperse, pathbuf_into_string, file_buffered) and the 2024 edition.

A nix development shell is available:

nix develop .#mia-dev

which provides the nightly toolchain, rust-analyzer and lldb.

Contributing

We DO NOT accept vibe coded contributions or contributions from agents.

However AI assisted code is fine as long as you the human are the one contributing.

Human contributions are always welcome.

Mia Agent

Mia is a coding and personal assistant AI agent designed to be unobtrusive, configurable and easy to use. Mia grows with you.

This project was inspired by Hermes Agent. Although it borrows ideas from it, this is not a direct derivation or modification of the project.

Principles

The agent should never write to the system without getting permission from the user.
Guardrails should be implemented in code, not in prompts.
The agent should act as an additional tool to the user not as a replacement for everything they already use.

Features

Minimal terminal UI powered by ratatui
Multi turn agent loop with configurable iteration limit
Agent memory (persistent across sessions)
Session history with auto-loading of the last session (mia session list / mia session clear)
Full markdown rendering in terminal (thanks to termimad)
Streaming responses (enabled by default, with token usage display)
OpenAI Compatible API access with support for multiple providers:
- OpenRouter (default, recommended)
- OpenCode Zen
- Google AI Studio
- Groq
- Cerebras
- Local LLMs (Ollama, LMStudio, llama.cpp, etc.)
Web search and extract using Tavily
Document conversion using pandoc (doc_convert tool)
Document creation from markdown/latex (doc_create tool)
Legacy reedline-based TUI available via the hidden mia old-tui command

Tools

Currently Mia has the following tools:

Tool	Description	Gated
📅	`datetime`	Get the current date and time
📁	`fs_list_dir`	List files in a directory
📖	`fs_read_file`	Read a file from the filesystem
🔍	`fs_grep_files`	Search file contents
🧭	`fs_search_dirs`	Search for files in a directory
✍️	`fs_write_file`	Write content to a file
🐍	`exec_python`	Execute Python 3 code
🐚	`exec_shell`	Run shell commands
🧠	`memory`	Manage memory about the user and the agent
🪏	`web_extract`	Extract content from a URL
🌐	`web_search`	Search the web
📄	`doc_create`	Create pdf, docx and other documents
📃	`doc_convert`	Convert any document to another using pandoc

Tools that are gated require explicit user permission before executing.

Installation

Option A: Prebuilt binaries

Prebuilt binaries for Linux (x86_64) and Windows (x86_64) are published as GitHub releases from the releases page.

Open a new terminal and run

mia setup

to configure your agent and start using it.

Option B: Using cargo (Windows/Linux)

Mia requires rust nightly

rustup toolchain install nightly
rustup default nightly

Cargo can fetch, build and install mia.

cargo +nightly install --git https://github.com/mastermach50/mia-agent

Run

mia setup

to configure your agent and start using it.

Option C: Using nix flakes (NixOS/Linux)

Run the given command to test out mia on your system.

nix run github:mastermach50/mia-agent

To install it, add to your system packages using the same flake.

Configuration

Mia supports multiple LLM providers:

OpenRouter (default, recommended) - can be obtained for free, there are also free models
OpenCode Zen - competitive pricing, free credits
Google AI Studio - free tier available
Groq - fast inference API
Cerebras - large model inference
Local LLMs - Ollama, LMStudio, llama.cpp, and other OpenAI-compatible local servers

Run mia setup to configure your agent interactively. mia setup supports section selection: --model, --tui and --agent to run only that part of the setup.
For advanced configuration, you can manually edit the ~/.mia/.env file on Linux or C:\Users\\.mia\.env on Windows to add your API key:

OPENROUTER_API_KEY=

Or for other providers, set the appropriate environment variable.

(Optional) Add your Tavily API key (can be obtained for free and has a reasonable number of free searches per month) to .mia/.env to enable the web_search and web_extract tools.

TAVILY_API_KEY=

(Optional) Run mia tools to see if all the tools are usable and find which components are missing if any.

Currently the external tools required by mia are

Usage

CLI

Use mia --help to access the full cli help menu.

TUI

The TUI supports certain commands. Use /help while in the TUI to see them.

Available Commands

mia tui - Start the main terminal UI (ratatui-based)
mia old-tui - Start the legacy reedline-based TUI (hidden command)
mia setup - Interactive setup wizard
mia tools - Check tool availability
mia model list - List available models
mia model list --free - List only free models (price filter)
mia model list --min-context - Filter by minimum context length (supports human readable values like 128k, 4M)
mia model list --max-price - Filter by maximum price per million tokens
mia model show - Show current model info
mia sessions list - List all sessions
mia session clear - Clear all sessions (prompts for confirmation)

Development

MSRV

Mia uses rust nightly because the code relies on unstable Rust features (iter_intersperse, pathbuf_into_string, file_buffered) and the 2024 edition.

A nix development shell is available:

nix develop .#mia-dev

which provides the nightly toolchain, rust-analyzer and lldb.

Contributing

We DO NOT accept vibe coded contributions or contributions from agents.

However AI assisted code is fine as long as you the human are the one contributing.

Human contributions are always welcome.

Mia Agent Mathew Manoj

README

Mia Agent

Table of Contents

Principles

Features

Tools

Installation

Option A: Prebuilt binaries

Option B: Using cargo (Windows/Linux)

Option C: Using nix flakes (NixOS/Linux)

Configuration

Usage

CLI

TUI

Available Commands

Development

MSRV

Contributing

Mia Agent Mathew Manoj

README

Mia Agent

Table of Contents

Principles

Features

Tools

Installation

Option A: Prebuilt binaries

Option B: Using cargo (Windows/Linux)

Option C: Using nix flakes (NixOS/Linux)

Configuration

Usage

CLI

TUI

Available Commands

Development

MSRV

Contributing