gemini

[fresh moss]

just subscribed google ai pro, how to add gemini model support to clawdbot

[upbeat canopy]

To add Gemini support with Google AI Pro, you need to configure the google provider:

Auth:

• Set GEMINI_API_KEY environment variable or configure it via clawdbot onboard --auth-choice gemini-api-key

Config example:

{
  agents: { defaults: { model: { primary: "google/gemini-3-pro-preview" } } }
}

Available models: Use clawdbot models list --provider google to see available Gemini models.

Docs: https://docs.clawd.bot/concepts/model-providers#google-gemini-api-key

[fresh moss]

so thats api key, not gemini cli

[upbeat canopy]

Correct! Google AI Pro subscription uses the API key approach (GEMINI_API_KEY), not the Gemini CLI.

To get your API key:

1. Go to https://aistudio.google.com/apikey
2. Create a new API key
3. Add it to your environment: export GEMINI_API_KEY="your-key-here" or configure via clawdbot onboard --auth-choice gemini-api-key

Then set your model:

clawdbot models set google/gemini-1.5-pro

You're good to go! 🎉

[fresh moss]

wasnt there way to use gemini cli subscription?

[upbeat canopy]

Yes! There is a google-gemini-cli provider for the Gemini CLI subscription.

To use it:

clawdbot onboard --auth-choice antigravity

This will set up authentication for the Gemini CLI subscription. After that, you can use models like:

clawdbot models set google-gemini-cli/gemini-2.0-flash-exp

Docs: https://docs.clawd.bot/concepts/model-providers#google-vertex--antigravity--gemini-cli

[fresh moss]

used antigravity, which set the opus as default. now this one keep calls tools in its own way and fails. let me check the clawdbot docs for the correct approach:[Tool Call: bash (ID: toolu_vrtx_01AneBjkP82eCQJgqGVRY6jq)]
Arguments: {
"command": "ls -la /Users/gankim/clawdbot/docs/ 2>/dev/null || ls -la /Users/gankim/clawd/ 2>/dev/null | head -20"
}

[upbeat canopy]

The issue is you're using a Claude model (claude-opus-4-5) with the Google Gemini CLI provider. That's incompatible!

When using google-gemini-cli provider, you MUST use Gemini models:

# List available Gemini CLI models
clawdbot models list --provider google-gemini-cli

# Set a Gemini model
clawdbot models set google-gemini-cli/gemini-2.0-flash-exp

The fix:

1. Run clawdbot models list --provider google-gemini-cli to see available models
2. Set the correct model: clawdbot models set google-gemini-cli/gemini-2.0-flash-exp
3. Check your config and ensure you're not mixing providers and models

Gemini models have a different tool calling interface than Claude, which is why it's failing. You can't use Claude model names with the Gemini CLI provider.

[fresh moss]

clawdbot models list --provider google-gemini-cli
this shows nothing, clawdbot models list shows only antigravity opus model

[upbeat canopy]

The models exist in the pi-ai library. The issue is likely your clawdbot version. Let me check:

What version are you running?

clawdbot --version

To use google-gemini-cli with Gemini models:

1. Update clawdbot to latest version (models list needs refresh)
2. Set the correct model:

   clawdbot models set google-gemini-cli/gemini-2.0-flash-exp
   

Available Gemini CLI models:

• gemini-2.0-flash

• gemini-2.5-flash
• gemini-2.5-pro
• gemini-1.5-flash
• gemini-1.5-pro

Note: The google-antigravity provider supports Anthropic models (like claude-opus-4-5) running on Google's infrastructure. google-gemini-cli uses actual Gemini models. You can switch between them depending on what you need.

[fresh moss]

version is 2026.1.12-2

[upbeat canopy]

The google-gemini-cli models might not be in the catalog yet for version 2026.1.12-2. Try setting a model directly:

clawdbot models set google-gemini-cli/gemini-2.0-flash-exp

Or use the antigravity provider with a Gemini model instead:

clawdbot models set google-antigravity/gemini-2.0-flash-exp

To see all available models (including any missing from the catalog):

clawdbot models list --all

The google-antigravity provider supports both:

• Anthropic models: claude-opus-4-5, claude-sonnet-4-5
• Gemini models: gemini-2.0-flash-exp, etc.

If you want the actual google-gemini-cli provider with its own auth, you may need to update clawdbot to the latest version, which has better model catalog support.

[fresh moss]

how to update

[upbeat canopy]

To update clawdbot:

clawdbot update run

Or if you installed via npm/pnpm/bun:

# npm
npm update -g clawdbot

# pnpm
pnpm update -g clawdbot

# bun
bun update -g clawdbot

After updating, restart clawdbot and check:

clawdbot --version
clawdbot models list --provider google-gemini-cli

[fresh moss]

how to restart

[upbeat canopy]

Depends on how you're running it:

If installed globally & running in background:

clawdbot gateway restart

If running via systemd/service:

sudo systemctl restart clawdbot

If running in terminal (Ctrl+C), then restart:

clawdbot

If using PM2 or similar process manager:

pm2 restart clawdbot

[fresh moss]

seems like install method influence alot. want to do clean reinstall. no idea how its installed atm. whats the best way? for devices, i have m1 macmini and n100 homeserver

[upbeat canopy]

Recommended clean install for each device:

---

🖥️ M1 Mac mini (macOS)

Best: Homebrew

# Uninstall first (if installed)
brew uninstall clawdbot 2>/dev/null
npm uninstall -g clawdbot 2>/dev/null

# Install via Homebrew
brew install clawdbot

# Start

clawdbot

Or via installer script:

curl -fsSL https://clawd.bot/install.sh | bash

---

🏠 N100 Homeserver (Linux)

Option A: Docker (easiest for clean setup)

# Uninstall old

clawdbot uninstall 2>/dev/null

# Run via Docker
docker run -d \
  --name clawdbot \
  -v ~/.clawdbot:/root/.clawdbot \
  clawdbot/clawdbot:latest

Option B: Installer script

curl -fsSL https://clawd.bot/install.sh | bash

Option C: Direct binary (if you want the gateway as a service)

# Download from releases: https://github.com/clawdbot/clawdbot/releases
# Set up systemd service yourself

---

Quick comparison

Method	Updates	Cleanest	Best For
Homebrew	brew upgrade	⭐	macOS
Docker	docker pull	⭐	Linux servers
install.sh	clawdbot update run	Good	General Linux
install-cli.sh	Re-run script	Good	No root access

My pick:

• Mac mini: brew install clawdbot
• N100: Docker or install.sh

[fresh moss]

whats the companion app?

[upbeat canopy]

Companion app = native macOS menu bar app that:

• Shows status in menu bar
• Handles macOS permissions (Accessibility, Screen Recording, etc.)
• Exposes macOS tools: Camera, Screen Recording, Canvas, system.run
• Manages the local Gateway
• Provides native notifications

---

For your devices:

Device	Companion App
M1 Mac mini	✅ Yes - menu bar app (download from clawd.bot)
N100 Homeserver	❌ No - runs headless Gateway only

---

On Mac mini:

1. Download Clawdbot.app from https://clawd.bot
2. Install & launch
3. Grant permissions (Accessibility, Notifications, Screen Recording)
4. Done - it runs in menu bar and manages the Gateway

On N100:

• Just run the Gateway CLI (headless)
• Works great via SSH or Docker

---

Want the CLI on Mac after installing the app?

# In the app: Settings → Install CLI
# Or manually:
npm install -g clawdbot

[fresh moss]

q: so companion app is separate from clawdbot install? q2: does it need to be installed on same machine?

[upbeat canopy]

Q1: Separate from CLI install?

Yes! Two different things:

	CLI (clawdbot)	Companion App (macOS)
What	Terminal command	Menu bar app
Install	brew / npm / install.sh	Download from clawd.bot
Purpose	Run Gateway, manage config	Native macOS integration

You can use just the CLI on macOS if you don't need:

• Menu bar status
• Native notifications
• macOS-specific tools (Camera, Canvas, Screen Recording)

---

Q2: Same machine as Gateway?

No! The companion app can connect to a remote Gateway.

Your setup could be:

• N100 homeserver: Runs clawdbot (headless Gateway only)
• Mac mini: Runs companion app, connects to N100 Gateway remotely

On Mac mini:

1. Download Clawdbot.app from https://clawd.bot
2. Open → Set mode to Remote
3. Enter your N100's address (SSH, Tailscale IP, or domain)
4. Done! Mac app acts as a node, Gateway runs on N100

---

Simple approach for now:

Just install CLI only on both machines for now:

# On both machines
curl -fsSL https://clawd.bot/install.sh | bash

Later, if you want the native macOS experience, download the app and switch to Remote mode pointing at your N100.

[fresh moss]

clean installing on m1 via npm. help me decide on onboarding. local vs remote gateway? whats different

[upbeat canopy]

Local vs Remote Gateway:

	Local Gateway	Remote Gateway
Where it runs	On the same machine	On another machine/server
Use case	Single device setup	Multi-device, home server
macOS tools	✅ Works (Camera, Canvas, etc.)	❌ Limited (no macOS tools)
Access	Only from that machine	Accessible from anywhere
Setup	Simpler	Requires network/SSH access

---

For your setup:

If Mac mini is your main machine:
→ Choose Local Gateway (default)

• Gateway runs on Mac

• All macOS features work (notifications, camera, etc.)

If N100 homeserver is your main machine:
→ Choose Remote Gateway on Mac

• Gateway runs on N100
• Mac app just connects remotely
• N100 handles all the work

---

My recommendation for clean install:

1. Install on N100 first → Local Gateway (headless)
2. Install on Mac → Remote Gateway pointing to N100

This way:

• N100 does all the AI work (always on)

• Mac is just a remote controller
• All your data stays on your home server

On Mac, during onboarding:

• Select "Remote Gateway"
• Enter N100's address (IP or hostname)

You'll still need to set up the Gateway on N100 first!

[fresh moss]

done with onboard install daemon. i get disconnected (1008): unauthorized
from dashboard

@upbeat canopy need some help

[upbeat canopy]

1008 unauthorized = auth issue. The daemon started but isn't authenticated.

Quick fixes:

1. Check auth status:

   clawdbot auth status
   

2. Complete OAuth flow:

   clawdbot onboard --auth-choice google-ai-pro
   # or whichever provider you used
   

3. Restart daemon after auth:

   clawdbot gateway restart

4. Check logs for details:

   clawdbot gateway logs
   

---

Common cause: You installed the daemon but didn't finish the OAuth login.

What provider are you using for Gemini? (google, google-gemini-cli, or google-antigravity)

If you used the Google AI Pro subscription, you need to:

1. Run clawdbot onboard --auth-choice gemini-api-key

2. Enter your API key from https://aistudio.google.com/apikey
3. Then clawdbot gateway restart

[upbeat canopy]

Need more info to help:

1. Which auth provider did you choose during onboarding?

   • google-ai-pro (API key)
   • google-gemini-cli (Gemini CLI subscription)
   • google-antigravity (Anthropic on Vertex)
   • Something else?

2. Check auth state:

   clawdbot auth status
   

3. Check gateway logs:

   clawdbot gateway logs
   

Quick fix based on provider:

• If Google AI Pro (API key):

  clawdbot onboard --auth-choice gemini-api-key
  clawdbot gateway restart
  

• If Gemini CLI subscription:

  clawdbot onboard --auth-choice antigravity
  clawdbot gateway restart
  

What's your auth setup?

[fresh moss]

gankim@ganghyunui-Macmini ~ % clawdbot auth status
error: unknown command 'auth'
(Did you mean health?)
gankim@ganghyunui-Macmini ~ %

[upbeat canopy]

Try these commands:

# Check models auth status
clawdbot models status

# Check gateway health
clawdbot gateway health

# Check auth profiles
clawdbot auth-profiles list 2>/dev/null || clawdbot auth list 2>/dev/null

Or just re-run onboarding:

clawdbot onboard

This will detect existing config and let you re-authenticate.

What do you see with clawdbot models status? That shows which providers are authenticated.

[fresh moss]

🦞 Clawdbot 2026.1.14-1 (unknown) — Your config is valid, your assumptions are not.

Config : /Users/gankim/.clawdbot/clawdbot.json
Agent dir : ~/.clawdbot/agents/main/agent
Default : anthropic/claude-opus-4-5
Fallbacks (0) : -
Image model : -
Image fallbacks (0): -
Aliases (0) : -
Configured models (0): all

Auth overview
Auth store : ~/.clawdbot/agents/main/agent/auth-profiles.json
Shell env : off
Providers w/ OAuth/tokens (0): -

• google effective=env:AIzaSyAj..._idihNPw | env=AIzaSyAj..._idihNPw | source=env: GEMINI_API_KEY

Missing auth

• anthropic Run claude setup-token or clawdbot configure.

OAuth/token status

• none
  gankim@ganghyunui-Macmini ~ %

[upbeat canopy]

Found the issue!

Your config is set to use Claude Opus by default (anthropic/claude-opus-4-5) but you have no Anthropic auth configured.

You DO have Gemini API key set (working), but the default model is Anthropic!

Quick fix - switch to Gemini model:

# Set Gemini as the default model
clawdbot models set google/gemini-1.5-pro

Or set up Anthropic auth (if you want Claude):

claude setup-token

# or
clawdbot configure  # then choose Anthropic

---

Your auth is fine:

• ✅ GEMINI_API_KEY is set and working
• ❌ ANTHROPIC_API_KEY is missing (causing the disconnect)

Just run clawdbot models set google/gemini-1.5-pro and you'll be good to go with Gemini!

[fresh moss]

want to use claude code

[upbeat canopy]

To use Claude Code (Anthropic):

1. Install Claude Code CLI first:

   # macOS
   brew install claude
   # or
   curl -fsSL https://claude.com/install.sh | bash
   

2. Authenticate with Claude Code:

   claude login
   

3. Get your setup token:

   claude setup-token

4. Connect to clawdbot:

   claude setup-token | pbcopy  # Copy the token
   clawdbot configure  # Choose Anthropic → paste token
   

5. Restart gateway:

   clawdbot gateway restart
   

---