fliiq 1.0.0

The AI agent that builds itself

Fliiq

The AI agent you actually own. One command to plan, build, and automate — with persistent memory and self-generating skills.

Quick Start

pip install fliiq
fliiq init                  # Creates ~/.fliiq/ + .env template
# Edit ~/.fliiq/.env with your API key (Anthropic, OpenAI, or Gemini)
fliiq                       # Interactive REPL from any terminal

Where does .env go?

• Pip-installed users: ~/.fliiq/.env (created by fliiq init) — this is the standard setup.
• Internal developers (source checkout): A project-root .env takes precedence over ~/.fliiq/.env. Fliiq checks for a local .env first, then falls back to the global one.

Project-specific setup (optional)

cd my-project
fliiq init --project        # Creates .fliiq/ with SOUL.md, playbooks/, mcp.json, etc.

Development install

git clone <repo-url> && cd Fliiq
pip install -e ".[dev]"
fliiq init

Features

Agent loop — Claude Code-style architecture where the model plans, executes, and evaluates in a single loop. Three modes: autonomous (full control), supervised (approve each tool call), and plan (plan only, then approve to execute).

26 core skills — File I/O (read_file, write_file, edit_file, list_directory), search (grep, find), system (shell, deps, dev_server), web (web_search, fetch_html), memory (memory_read, memory_write, memory_search), email (send_email, receive_emails, mark_email_read, delete_email, archive_email), SMS (send_sms, receive_sms), messaging (send_telegram), calendar (google_calendar), contacts (find_email), time (get_current_time), and music (spotify).

Persistent memory — Curated MEMORY.md loaded every session. Daily logs, skill-specific memories, and keyword search across all memory files. The agent reads and updates memory naturally.

Self-generating skills — When the agent detects a capability gap, it researches the API, generates a complete skill (SKILL.md + fliiq.yaml + main.py), installs it, and uses it immediately. Skills persist across sessions in .fliiq/skills/.

Daemon and jobs — Background process that runs scheduled tasks. Cron, interval, one-shot, and webhook triggers. Each job gets its own memory file and audit trail. Create jobs via CLI or let the agent create them mid-conversation.

Customizable identity — SOUL.md defines the agent's personality. Playbooks add domain-specific instructions. Both are scaffolded from templates and fully editable per-project.

Email, SMS, and Telegram — Send and receive email (Gmail OAuth or app password), SMS (Twilio), and Telegram messages. Two modes: Fliiq's own channels (people message the bot) and managing your accounts (the agent reads your inbox, sends on your behalf).

Google Calendar — Full calendar management via OAuth. List calendars, create/update/delete events, check availability. Multi-account support — authorize multiple Google accounts.

MCP support — Connect any MCP server (stdio or streamable-http) and its tools are available to the agent alongside built-in skills. One command to add, test, and manage servers.

Full-screen TUI — Textual-based interface with message scrolling, mode indicator, thinking timer, and keyboard shortcuts (fliiq tui).

Usage

# Interactive (default)
fliiq                               # REPL (autonomous mode)
fliiq --mode supervised             # REPL in supervised mode
fliiq tui                           # Full-screen TUI

# Single-shot
fliiq run "build a Flask todo app"
fliiq run "what time is it" --mode autonomous
fliiq plan "refactor the auth module"

# Skills
fliiq skill-list                    # Show all skills with source (core/local)
fliiq skill-promote <name>          # Promote local skill to core

# Identity and customization
fliiq soul show                     # Display SOUL.md (default + overrides)
fliiq soul edit                     # Open .fliiq/SOUL.md in $EDITOR
fliiq soul reset                    # Remove overrides, revert to defaults
fliiq playbook list                 # List playbooks with source (custom/bundled)
fliiq playbook show coding          # Display a playbook
fliiq playbook create devops        # Scaffold a new custom playbook

# Google accounts
fliiq google auth                   # Authorize a Google account (Calendar + Gmail)
fliiq google accounts               # List authorized accounts

# MCP servers
fliiq mcp add github --command npx --args "@modelcontextprotocol/server-github"
fliiq mcp add my-api --url https://mcp.example.com/mcp
fliiq mcp list                      # Show configured servers
fliiq mcp test                      # Validate connections
fliiq mcp remove github             # Remove a server

# Daemon and jobs
fliiq daemon start                  # Start background daemon
fliiq daemon start --detach         # Detach to background
fliiq daemon status                 # Check if running
fliiq daemon stop                   # Stop daemon

fliiq job list                      # List all jobs
fliiq job create                    # Create a job interactively
fliiq job run <name>                # Manual trigger
fliiq job logs <name>               # View run history
fliiq job delete <name>             # Remove a job

Chat Commands

Inside the REPL (fliiq):

Command	Action
/mode	Cycle mode (plan -> supervised -> autonomous)
/status	Show session info
/clear	Reset conversation history
/help	Show available commands
/exit	Exit chat

Architecture

fliiq/
  cli/          # Typer CLI, Rich display, Textual TUI, REPL
  runtime/
    agent/      # Agent loop, tool registry, prompt assembly, audit
    llm/        # LLM providers (Anthropic, OpenAI, Gemini) + failover
    skills/     # Skill loader, base class, installer
    mcp/        # MCP client, server connections, tool forwarding
    planning/   # Domain detection, playbook loading, reflection
    memory/     # Memory manager, keyword retrieval
    scheduler/  # Job executor, scheduler, run logging
  api/          # FastAPI daemon, webhook receiver
  data/         # Bundled skills, SOUL.md, playbooks, templates

Agent loop (runtime/agent/loop.py): Model calls tools, tools return results, loop continues until the model stops or hits max iterations. Mode enforcement filters available tools and gates execution.

Skill system (runtime/skills/): Each skill is a directory with SKILL.md (metadata), fliiq.yaml (schema), and main.py (async handler). Discovery scans bundled skills first, then project-level overrides, then user-local .fliiq/skills/.

Memory (runtime/memory/): Files in .fliiq/memory/. MEMORY.md (curated, always in prompt), daily logs (YYYY-MM-DD.md), skill memories (skills/*.md). Agent reads/writes via memory skills.

Configuration

API Keys (.env)

At least one required:

ANTHROPIC_API_KEY=your-key
OPENAI_API_KEY=your-key
GEMINI_API_KEY=your-key

Priority: Anthropic > OpenAI > Gemini.

Global Directory (~/.fliiq/)

Created by fliiq init — used from any terminal:

~/.fliiq/
  .env          # API keys + Fliiq channel credentials
  user.yaml     # Your identity (name, emails, timezone)
  memory/       # Persistent memory files
  audit/        # Audit trails from agent runs
  skills/       # User-generated skills (available everywhere)

Project Directory (.fliiq/) — optional

Created by fliiq init --project — overrides global for this project:

.fliiq/
  SOUL.md       # Agent personality overrides (scaffolded from template)
  playbooks/    # Custom domain playbooks
  mcp.json      # MCP server connections
  memory/       # Project-specific memory
  audit/        # Project audit trails
  jobs/         # Scheduled job definitions (YAML)
  skills/       # Project-specific skills

Resolution: local .fliiq/ > global ~/.fliiq/ > bundled defaults.

Agent Identity (SOUL.md)

SOUL.md defines the agent's personality, communication style, and behavioral rules. A bundled default ships with the package. To customize for your project:

fliiq soul edit             # Opens .fliiq/SOUL.md in $EDITOR (creates from template if missing)
fliiq soul show             # See default + your overrides
fliiq soul reset            # Delete overrides, revert to bundled default

Your .fliiq/SOUL.md is appended to the default as "User Overrides" — you only need to specify what you want to change. The bundled default is never modified.

Custom Playbooks

Playbooks are domain-specific instructions loaded when the domain detector matches keywords in your prompt. A bundled coding playbook activates for coding tasks.

fliiq playbook create devops    # Scaffold .fliiq/playbooks/devops.md from template

Edit the playbook and add trigger keywords on the # Keywords: line:

# Keywords: devops, deploy, kubernetes, terraform, infrastructure, ci/cd

Fliiq matches keywords in your prompt against these. When 2+ keywords match, the playbook loads into the agent's system prompt. fliiq playbook list shows all playbooks with their source (custom/bundled).

MCP Servers

Connect external tools via the Model Context Protocol. Fliiq supports stdio and streamable-http transports.

# Add a stdio server (e.g. from npm)
fliiq mcp add github --command npx --args "@modelcontextprotocol/server-github"

# Add a streamable-http server
fliiq mcp add my-api --url https://mcp.example.com/mcp

# Verify it works
fliiq mcp test github

# List / remove
fliiq mcp list
fliiq mcp remove github

MCP tools are auto-discovered at startup and registered alongside built-in skills. The agent sees them as regular tools (prefixed with mcp_{server}_{tool}). Connection failures are non-fatal — other servers and skills still work.

Config is stored in .fliiq/mcp.json (scaffolded by fliiq init --project). You can also edit it directly:

{
  "servers": {
    "github": {
      "command": "npx",
      "args": ["@modelcontextprotocol/server-github"],
      "transport": "stdio"
    }
  }
}

User Profile (~/.fliiq/user.yaml)

Your identity file — loaded into every agent session so the agent knows who you are. When you say "my email" or "my calendar," the agent uses these accounts.

name: John Doe
emails:
  - address: [e-mail]
    label: personal
  - address: [e-mail]
    label: work
timezone: America/New_York

Created by fliiq init. Emails are auto-added when you run fliiq google auth.

Google Account Integration

Authorize your Google accounts so the agent can manage your Gmail and Calendar:

fliiq google auth           # Opens browser for OAuth consent (Calendar + Gmail)
fliiq google accounts       # List authorized accounts

Run fliiq google auth once per Google account. Each gets OAuth tokens stored in ~/.fliiq/google_tokens.json. After auth, you'll be prompted to add the email to your user profile.

Prerequisites:

1. Create a project at Google Cloud Console
2. Enable Google Calendar API and Gmail API
3. Create OAuth 2.0 credentials (Desktop app, redirect URI http://localhost:8080/callback)
4. Add these to your .env:

GOOGLE_CLIENT_ID=your-client-id
GOOGLE_CLIENT_SECRET=your-client-secret

Then ask the agent things like "check my email," "book a meeting tomorrow at 3pm," or "find my availability this week."

Communication Channels

Fliiq supports two distinct use cases for email, SMS, and messaging:

Use Case 1: Fliiq's own channels (talk TO Fliiq)

Give Fliiq its own email, phone number, and Telegram bot so people can message it directly — like texting an assistant.

# Fliiq's bot email (receives messages on behalf of Fliiq)
FLIIQ_GMAIL_ADDRESS=[e-mail]
FLIIQ_GMAIL_APP_PASSWORD=your-app-password

# Fliiq's phone number (Twilio)
TWILIO_ACCOUNT_SID=your-sid
TWILIO_AUTH_TOKEN=your-token
TWILIO_PHONE_NUMBER=[phone]

# Fliiq's Telegram bot
TELEGRAM_BOT_TOKEN=your-bot-token
TELEGRAM_ALLOWED_CHAT_IDS=123456789

These are Fliiq's OWN inboxes. The daemon monitors them for inbound messages and responds automatically.

Telegram Setup

1. Create a bot via @BotFather on Telegram and copy the bot token

2. Add TELEGRAM_BOT_TOKEN=your-bot-token to ~/.fliiq/.env

3. Run the interactive setup to detect your chat ID:

   fliiq telegram setup
   

   This will prompt you to send a message to your bot, then auto-detect and save your chat ID.

   Manual alternative: Add TELEGRAM_ALLOWED_CHAT_IDS=<chat_id> to ~/.fliiq/.env directly. Comma-separate multiple IDs.

Required: When TELEGRAM_BOT_TOKEN is set, TELEGRAM_ALLOWED_CHAT_IDS must also be set. The daemon will refuse to start without it — this prevents unauthorized users from accessing your agent.

Use Case 2: Your accounts (Fliiq manages FOR you)

Authorize your personal/work Google accounts so Fliiq can read your email, send on your behalf, and manage your calendar.

fliiq google auth           # Authorize each Google account via OAuth

Emails go in ~/.fliiq/user.yaml, and the agent uses OAuth (not app passwords) to access them. This is what powers "check my email" and "schedule a meeting."

The two use cases are independent. You can set up one, both, or neither. The agent's system prompt distinguishes between "Fliiq's email" (from .env) and "the user's email" (from user.yaml).

Troubleshooting

fliiq doctor                # Verify setup: API keys, SOUL.md, playbooks, MCP, skills

Security

Fliiq gives an LLM agent access to your filesystem, email, SMS, Telegram, and shell commands. That power is the point — but it comes with real risks.

What Fliiq protects

• Credential file deny list — The agent cannot read or write ~/.fliiq/.env, ~/.fliiq/google_tokens.json, ~/.fliiq/daemon.secret, or anything in ~/.ssh/, ~/.aws/, ~/.gnupg/. This prevents prompt injection attacks from exfiltrating secrets.
• Prompt injection defense — All inbound external content (Telegram messages, emails, SMS, webhook payloads) is wrapped in <external_message> tags with a system prompt instruction telling the agent to never follow instructions from external sources.
• Telegram allowlist — TELEGRAM_ALLOWED_CHAT_IDS is required when a bot token is set. Unauthorized users get a hardcoded rejection reply — no LLM call, no tool access.
• Daemon API authentication — All /api/* routes require a Bearer token (auto-generated at ~/.fliiq/daemon.secret). Prevents local CSRF and rogue processes from triggering agent execution.
• Package install validation — The deps skill validates package names against a regex and uses subprocess_exec (no shell) to prevent command injection.

What Fliiq does NOT protect

• Your project files — The agent has full read/write access to your working directory. This is by design (it needs to edit your code), but a prompt injection attack could modify or delete project files.
• Self-corruption — Fliiq can overwrite its own local configuration (~/.fliiq/jobs/, ~/.fliiq/user.yaml, skill files, etc.). If the agent corrupts its local state, reset with:

  rm -rf ~/.fliiq && fliiq init
  

  This is safe — core package code lives in site-packages/ (read-only via pip install). Only local config and job definitions are lost.
• System prompt extraction — An attacker with access to the agent can extract the system prompt. This is a soft defense only (LLMs can be jailbroken).
• Audit log contents — Audit logs in ~/.fliiq/audit/ may contain sensitive conversation data.

Best practices

1. Use supervised mode for untrusted tasks — fliiq run "..." --mode supervised requires your approval before each tool call.
2. Review scheduled jobs — Jobs run autonomously in the daemon. Audit ~/.fliiq/jobs/ to know what's running.
3. Don't put secrets in prompts — The agent resolves credentials from env vars and OAuth tokens automatically. Never include passwords in job prompts or Telegram messages.
4. Back up your project — Use git. The agent writes files. Commits give you rollback.
5. Rotate daemon secret after exposure — Delete ~/.fliiq/daemon.secret and restart the daemon to regenerate.

Development

# Install with dev dependencies
pip install -e ".[dev]"

# Lint
ruff check fliiq/

# Tests
pytest tests/

# Run specific test file
pytest tests/test_agent_loop.py -v

Tech Stack

• Python 3.12+, Typer (CLI), Rich (display), Textual (TUI)
• FastAPI + uvicorn (daemon), croniter (scheduling)
• Anthropic / OpenAI / Gemini SDKs (LLM providers)
• structlog (logging), pytest (testing), ruff (linting)

License

TBD