ask-gemini 0.2.1

CLI tool to chat with Gemini from the terminal — no API key needed

ask-gemini

CLI tool to chat with Gemini's web interface from the terminal. No API key needed — uses your browser cookies.

For agents: See SKILL.md for agent usage patterns and best practices. Quick install: pip install ask-gemini (PyPI) or download from GitHub Releases.

Setup

Option A: Install via pip (recommended)

pip install ask-gemini

Available on PyPI.

Option B: Download standalone executable

Download the latest macOS executable from GitHub Releases:

# Download from the latest release asset at https://github.com/Guitenbay/ask-gemini/releases
chmod +x ask-gemini
./ask-gemini "Hello!"

No Python installation required.

Option C: Build from source

git clone https://github.com/Guitenbay/ask-gemini.git
cd ask-gemini
pip install pyinstaller
bash build.sh

The executable is at dist/ask-gemini (~21MB). Or install as a Python package:

pip install -e .

Configure cookies

Option A: Auto-detect from browser

If you're logged into Gemini in Chrome or Edge, ask-gemini will auto-detect your cookies. Just run:

ask-gemini "Hello!"

Option B: Manual setup

ask-gemini --cookie-setup

Follow the instructions to copy __Secure-1PSID and __Secure-1PSIDTS from your browser, then save them in a .env file.

Usage

Single question

ask-gemini "What is Python?"
ask-gemini --model gemini-3-flash "Explain async/await"
ask-gemini --no-stream "Write a haiku about coding"

Chat mode

Enter an interactive conversation that resumes your latest web-side chat:

ask-gemini --chat

This automatically picks up where you left off on gemini.google.com. Type your message, and Gemini responds. Useful commands while in chat:

Command	Action
exit / quit	Leave chat mode
new	Start a fresh conversation
clear	Same as new

To always start a fresh conversation instead of resuming:

ask-gemini --chat --new-chat

Chat mode internals

• The same cid (conversation ID) is used as the web interface, so history is shared between CLI and browser
• Multiple CLI sessions in a row will keep using the same web conversation
• Switching models in an existing chat starts a new session on the web side

Named sessions

Create isolated conversations that don't interfere with each other. Perfect for multiple agents working on different tasks:

# Create or resume a named session
ask-gemini --session code-review "Analyze this code: $(cat app.py)"

# Follow-up — Gemini remembers the code review context
ask-gemini --session code-review "Now rewrite it with better error handling"

# A completely separate session
ask-gemini --session data-analysis "What is SQL?"

Manage your sessions:

ask-gemini --sessions                # List all saved sessions
ask-gemini --rm-session code-review  # Delete a session

Sessions are stored in ~/.ask-gemini/sessions.json and map directly to conversations on gemini.google.com.

Options

Option	Description
-m, --model	Model to use (default: gemini-3-pro)
-s, --stream	Stream output as it arrives (default)
--no-stream	Wait for full response before printing
--chat	Enter interactive chat mode (resumes web conversation)
--new-chat	Start a fresh conversation instead of resuming
--session <name>	Use a named, isolated conversation session
--sessions	List all saved named sessions
--rm-session <name>	Delete a named session
--no-rate-limit	Disable human-like typing/cooldown delays
--cookie-setup	Print cookie setup instructions
--version	Print version

Available Models

Model	Description
gemini-3-pro	Most capable (default)
gemini-3-flash	Fast
gemini-3-flash-thinking	Flash with chain-of-thought

Configuration

All configuration via environment variables in .env (project root or ~/.ask-gemini/.env):

Variable	Description	Default
GEMINI_PSID	__Secure-1PSID cookie	(auto-detect from browser)
GEMINI_PSIDTS	__Secure-1PSIDTS cookie	(auto-detect from browser)
PROXY_URL	HTTP proxy URL	(empty)
MODEL	Default model	gemini-3-pro

Rate limiting

By default, ask-gemini simulates human typing speed to avoid triggering Google's risk control:

• Typing delay: message_length × TYPING_SPEED seconds, clamped between MIN_DELAY and MAX_DELAY
• Cooldown: COOLDOWN seconds after receiving a response before the next request
• Jitter: +/- 20% random variation so delays aren't perfectly uniform

Configure in .env:

Variable	Description	Default
RATE_LIMIT	Enable/disable rate limiting (1 or 0)	1
TYPING_SPEED	Seconds per character (≈ 20 chars/sec)	0.05
MIN_DELAY	Minimum delay before sending (seconds)	1.0
MAX_DELAY	Maximum delay cap (seconds)	5.0
COOLDOWN	Seconds to wait after receiving a response	2.0

Or disable per-call with --no-rate-limit.

Notes

• Uses the free Gemini web interface, not the paid API.
• Cookies expire. If requests fail, refresh Gemini in your browser or update cookies manually.
• Automated access may violate Google's Terms of Service. Use at your own risk.