DOCUMENTATION

Learn Quill-Agent

Complete guide to installing, configuring, and using Quill-Agent. The self-improving AI agent that learns and improves with every task.

Get Started GitHub

Installation

Install Quill-Agent in 60 seconds on any platform. Works on Linux, macOS, Windows, WSL2, and even Android via Termux.

Linux / macOS / WSL2

bash

curl -fsSL https://raw.githubusercontent.com/quill-agent/quill-agent/main/scripts/install.sh | bash

Windows (PowerShell)

powershell

irm https://raw.githubusercontent.com/quill-agent/quill-agent/main/scripts/install.ps1 | iex

Homebrew (macOS)

bash

brew tap quill-agent/tap
brew install quill-agent

Android (Termux) — Use the same curl one-liner as Linux. The installer auto-detects Termux and configures accordingly.

Quickstart

Get up and running with your first conversation in under 2 minutes.

1. Initialize Quill-Agent

bash

quill-agent init --model llama3:8b

This creates your config file and sets up the default model. You can use any model from Ollama, OpenAI, Anthropic, or other providers.

2. Start the agent

bash

quill-agent start

3. Chat with your agent

The CLI will open an interactive session. Try asking it to:

•"Create a Python script that monitors CPU usage"
•"Research the top 3 competitors in my market and summarize"
•"Set up a cron job to check my website uptime every hour"

Learning Path

Find the right docs for your experience level.

New to Quill-Agent?

Start with Quickstart, then explore Tools and Memory.

Setting up for production?

Read Configuration, Security, and Architecture.

Want to extend capabilities?

Check out Skills System and MCP Integration.

Connecting from mobile?

Set up Messaging Gateway for Telegram, Discord, etc.

Configuration

Quill-Agent uses a YAML config file located at ~/.quill-agent/config.yaml

Basic Configuration

yaml

# Model configuration
model:
  provider: ollama        # ollama, openai, anthropic, etc.
  name: llama3:8b
  temperature: 0.7

# Memory settings
memory:
  enabled: true
  auto_persist: true
  
# Tools
tools:
  enabled:
    - web_search
    - file_operations
    - code_execution
    - shell_commands

Supported Providers

• Ollama — Local models, free, private
• OpenAI — GPT-4, GPT-3.5
• Anthropic — Claude 3, Claude 2
• OpenRouter — Access to 100+ models
• Groq — Ultra-fast inference

Messaging Gateway

Control Quill-Agent from anywhere — your phone, laptop, or any device. 20+ platforms supported.

TelegramStable

DiscordStable

SlackStable

WhatsAppBeta

SignalBeta

MatrixStable

EmailStable

SMSBeta

Microsoft TeamsAlpha

Setup Example (Telegram)

bash

quill-agent auth telegram
# Follow the prompts to connect your Telegram bot

Tools & Toolsets

70+ built-in tools organized into toolsets. Enable what you need.

Web Tools

web_search, web_browse, web_extract, screenshot

File Tools

read_file, write_file, list_directory, file_search

Code Tools

execute_code, run_shell, git_operations

Data Tools

json_parse, csv_process, database_query

Media Tools

image_generate, tts_speak, vision_analyze

Memory System

Quill-Agent remembers everything across sessions. It builds a persistent model of your preferences, projects, and context.

FTS5 Full-Text Search

All memories are indexed with SQLite FTS5 for instant cross-session recall. LLM summarization keeps context relevant.

Auto-Persist

The agent periodically nudges itself to persist important knowledge. No manual memory management needed.

User Model

Builds a deepening understanding of who you are, your preferences, communication style, and workflows.

Skills System

Procedural memory that the agent creates, improves, and reuses. Skills are portable and shareable.

Auto-Creation

When the agent solves a novel problem, it can create a skill for future reuse.

Self-Improvement

Skills improve during use — the agent refines them based on feedback and outcomes.

Skills Hub

Community-contributed skills. Import pre-built capabilities with a single command.

MCP Integration

Connect to Model Context Protocol servers to extend Quill-Agent's capabilities safely.

yaml

# config.yaml
mcp:
  servers:
    - name: filesystem
      command: npx @modelcontextprotocol/server-filesystem
      args: ["/home/user/projects"]
    - name: github
      command: npx @modelcontextprotocol/server-github
      env:
        GITHUB_TOKEN: $GITHUB_TOKEN

Voice Mode

Real-time voice interaction in CLI, Telegram, and Discord.

bash

quill-agent start --voice

Requires a microphone for input. TTS output works with any audio device. Supports Whisper for transcription and various TTS providers.

Security

Multiple layers of protection for safe autonomous operation.

Command Approval

Require human approval for sensitive operations. Configure per-tool or globally.

Container Isolation

Run code execution in isolated Docker containers with resource limits.

Authorization

Allowlist specific users across messaging platforms. Restrict access to sensitive tools.

Architecture

How Quill-Agent works under the hood.

Core Loop

Observe → Think → Act → Learn. The agent continuously cycles through perception, reasoning, action, and memory updates.

Terminal Backends

6 backends: Local, Docker, SSH, Daytona, Singularity, Modal. Choose based on your isolation and persistence needs.

Subagent Delegation

Spawn isolated subagents for parallel workstreams. Each has its own context but shares memory.

FAQ & Troubleshooting

How is this different from ChatGPT?

Quill-Agent runs locally, learns from experience, executes real actions, and doesn't forget you between sessions. It's an autonomous agent, not a chatbot.

Do I need a GPU?

No. CPU inference works fine with smaller models. For better performance, a GPU helps but isn't required.

Is my data sent anywhere?

Only if you configure a cloud LLM provider. With Ollama, everything stays local. You own your data.

Can I use this for work?

Yes, it's MIT licensed. Use it however you want, commercially or personally.

How do I update?

Run the same install command — it handles updates automatically.