AI with FZ
25 min video
3 min read
Free Claude Desktop Forever: Multi-Model Setup Guide
You just saved 22 min.
The big takeaway
Use OmniRoute to aggregate 258+ free API providers (Claude, Nvidia, OpenAI, DeepSeek) into Claude Desktop via gateway configuration, eliminating all subscription costs. Requires terminal setup, AntiGravity authentication, and model combo creation.
The Free Claude Desktop Opportunity
Claude Desktop is now completely free
As of July 16th, Claude Desktop no longer requires any payment. By integrating free API providers through OmniRoute, users can access all Claude models (Opus, Sonnet, Haiku) plus models from Nvidia, OpenAI, and DeepSeek without spending money.
0 rupees
Cost for Claude Desktop
Permanent free access to all integrated models
Multiple model providers available simultaneously
Within a single Claude Desktop instance, you can run models from four major providers: AntiGravity (Claude), Nvidia, OpenAI (open-source), and others. Each provider offers free tier models that can be tested and deployed.
1
AntiGravity (Claude)
~9 free models
2
Nvidia
11+ free models
3
OpenAI (open-source)
Multiple free models
4
AWS (Bedrock)
Free credits available
Free models across four major providers
OmniRoute: The Aggregation Engine
OmniRoute provides 258+ API provider integrations
OmniRoute is a gateway platform that centralizes access to 258+ different AI model providers, with 90+ offering free API access. It serves as the middleware between Claude Desktop and various free model sources.
258+
Available providers in OmniRoute
90+ providers offer free API access
Installation requires three terminal commands
OmniRoute is installed via npm with a single command (npm install -g omniroot), then launched locally. The setup opens a web UI at localhost where you configure provider connections and create model combos.
1
Run: npm install -g omniroot
2
Launch OmniRoute UI locally
3
Authenticate with provider accounts
4
Import free models
5
Create model combos
6
Configure Claude Desktop gateway
Six-step setup from terminal to working Claude Desktop
Connecting Claude Models via AntiGravity
AntiGravity provides free Claude model access
AntiGravity is Claude's official free tier provider. Connect via Gmail authentication in OmniRoute, then auto-sync and import available models. Nine free Claude models become available: Opus 4.6, Sonnet 4.6, Haiku 3.1, and others.
1
Claude Opus 4.6
Free
2
Claude Sonnet 4.6
Free
3
Claude Haiku 3.1
Free
4
Claude 3.1 Pro
Free
Four major Claude models available free via AntiGravity
Model combos route requests intelligently
Create named combos (e.g., 'Auto-Claude-Haiku') that group multiple models with either priority or round-robin logic. Priority calls the first model, then falls back if it fails; round-robin distributes load across models.
Priority mode
Try Model 1 → fallback to Model 2
Round-robin mode
Distribute requests across all models equally
Two routing strategies for model combos
Integrating Nvidia Free Models
Nvidia offers free API access with account creation
Create a free Nvidia account, generate an API key, and paste it into OmniRoute. Nvidia's free tier includes models like Nemotron, Mistral, and Mixtral, all available without payment.
1
Nemotron
Free
2
Mistral
Free
3
Mixtral 2.7
Free
4
DeepSeek Flash
Free
Nvidia's free model lineup
Nvidia model responses are slower than Claude
Nvidia models take noticeably longer to respond compared to Claude's models. This is a known limitation but not a blocker—responses still arrive reliably, just with higher latency.
Adding OpenAI Open-Source Models
OpenAI open-source models integrate with single step
OpenAI's open-source models (like GPT-4 variants) are available through OmniRoute with minimal setup. Delete any existing integration, add a new model connection, and import the free catalog.
Claude Desktop Configuration
Enable Developer Mode to access gateway settings
In Claude Desktop, click the three-dot menu, go to Help > Troubleshooting > Enable Developer Mode. This unlocks the Developer menu where you configure third-party API gateways.
1
Click three-dot menu in Claude Desktop
2
Select Help → Troubleshooting
3
Click Enable Developer Mode
4
Restart Claude Desktop
5
Click three dots again
6
Select Developer → Configure Third-Party Interface
Steps to unlock gateway configuration
Gateway configuration requires base URL and API key
In the Configure Third-Party Interface section, paste your OmniRoute endpoint as the base URL, select 'static API key' as the auth type, and paste your OmniRoute-generated API key. Enable model discovery to test the connection.
1
Copy OmniRoute endpoint URL
2
Paste as Base URL in Claude Desktop
3
Generate API key in OmniRoute
4
Paste API key in Claude Desktop
5
Select 'static API key' auth type
6
Enable model discovery and test
Gateway configuration workflow
Model names must match combo names exactly
The model names you add in Claude Desktop's gateway configuration must exactly match the combo names created in OmniRoute. For example, if you created a combo named 'Auto-Claude-Haiku', add a model with that exact name in Claude Desktop.
Model discovery often fails; manual model entry is required
Enabling model discovery typically returns errors. Instead, manually add each model by copying the combo names from OmniRoute and entering them individually in Claude Desktop's model configuration.
Troubleshooting and Optimization
First API calls take longer due to initialization
The first request to any model (especially Nvidia) will have higher latency as the service initializes. Subsequent requests are faster. This is normal and expected.
Model sequence priority can cause failures
If a model combo fails, edit it and reorder the model sequence, moving more reliable models higher in priority. Switch from priority mode to round-robin if issues persist.
Logs in OmniRoute show real-time request details
Right-click the OmniRoute combo and open its logs in a new tab to see which model was called, response times, and any errors. This helps debug routing issues.
Use ChatGPT to extract model names for manual entry
When model discovery fails, copy all available models from OmniRoute into ChatGPT and ask it to list all model names. Use this list to manually add models to Claude Desktop.
Real-World Usage
Claude Desktop can now access project folders and files
With the gateway configured, Claude Desktop can analyze files in your local directories. Select a folder, and Claude will use whichever model combo you've configured to process documents and answer questions about them.
No subscription limits or token restrictions
By aggregating free providers, you bypass individual provider rate limits. If one provider's free tier is exhausted, another combo automatically handles the request.
Worth quoting
"From now on, you won't have to pay a single rupee for Claude Desktop."
— AI with FZ, at [0:00]
"Inside one Claude, all original Claude models are running. NVIDIA's free models are running. OpenAI's free models are running."
— AI with FZ, at [1:35]
"It took me four days to debunk and develop it. If it works for you, please say in the comments."
— AI with FZ, at [24:06]
Try this
Install OmniRoute globally using: npm install -g omniroot
Launch OmniRoute and authenticate with Gmail to connect AntiGravity (Claude models)
Create a free Nvidia account and generate an API key, then add it to OmniRoute
Create model combos in OmniRoute for each provider (e.g., Auto-Claude-Haiku, Auto-Nvidia-Nemotron)
Uninstall Claude Desktop completely, then download and reinstall from claude.ai
Enable Developer Mode in Claude Desktop (Help > Troubleshooting > Enable Developer Mode)
Configure the gateway: paste OmniRoute endpoint as base URL and API key in Developer settings
Manually add each model combo name to Claude Desktop's model configuration
Test each model combo by sending a message and verifying the response in OmniRoute logs
Adjust model sequence or switch to round-robin mode if any combo fails consistently
Made with Glimpse by Wozart
glimpse.wozart.com/v/8x4hlb6u
Share this infographic
Read this infographic as text

Free Claude Desktop Forever: Multi-Model Setup Guide

Summary of the video “Claude Desktop Is Now FREE?! Use ALL Claude Models (Full Setup) by AI with FZ.

Use OmniRoute to aggregate 258+ free API providers (Claude, Nvidia, OpenAI, DeepSeek) into Claude Desktop via gateway configuration, eliminating all subscription costs. Requires terminal setup, AntiGravity authentication, and model combo creation.

The Free Claude Desktop Opportunity

Claude Desktop is now completely free

As of July 16th, Claude Desktop no longer requires any payment. By integrating free API providers through OmniRoute, users can access all Claude models (Opus, Sonnet, Haiku) plus models from Nvidia, OpenAI, and DeepSeek without spending money.

Multiple model providers available simultaneously

Within a single Claude Desktop instance, you can run models from four major providers: AntiGravity (Claude), Nvidia, OpenAI (open-source), and others. Each provider offers free tier models that can be tested and deployed.

OmniRoute: The Aggregation Engine

OmniRoute provides 258+ API provider integrations

OmniRoute is a gateway platform that centralizes access to 258+ different AI model providers, with 90+ offering free API access. It serves as the middleware between Claude Desktop and various free model sources.

Installation requires three terminal commands

OmniRoute is installed via npm with a single command (npm install -g omniroot), then launched locally. The setup opens a web UI at localhost where you configure provider connections and create model combos.

Connecting Claude Models via AntiGravity

AntiGravity provides free Claude model access

AntiGravity is Claude's official free tier provider. Connect via Gmail authentication in OmniRoute, then auto-sync and import available models. Nine free Claude models become available: Opus 4.6, Sonnet 4.6, Haiku 3.1, and others.

Model combos route requests intelligently

Create named combos (e.g., 'Auto-Claude-Haiku') that group multiple models with either priority or round-robin logic. Priority calls the first model, then falls back if it fails; round-robin distributes load across models.

Integrating Nvidia Free Models

Nvidia offers free API access with account creation

Create a free Nvidia account, generate an API key, and paste it into OmniRoute. Nvidia's free tier includes models like Nemotron, Mistral, and Mixtral, all available without payment.

Nvidia model responses are slower than Claude

Nvidia models take noticeably longer to respond compared to Claude's models. This is a known limitation but not a blocker—responses still arrive reliably, just with higher latency.

Adding OpenAI Open-Source Models

OpenAI open-source models integrate with single step

OpenAI's open-source models (like GPT-4 variants) are available through OmniRoute with minimal setup. Delete any existing integration, add a new model connection, and import the free catalog.

Claude Desktop Configuration

Enable Developer Mode to access gateway settings

In Claude Desktop, click the three-dot menu, go to Help > Troubleshooting > Enable Developer Mode. This unlocks the Developer menu where you configure third-party API gateways.

Gateway configuration requires base URL and API key

In the Configure Third-Party Interface section, paste your OmniRoute endpoint as the base URL, select 'static API key' as the auth type, and paste your OmniRoute-generated API key. Enable model discovery to test the connection.

Model names must match combo names exactly

The model names you add in Claude Desktop's gateway configuration must exactly match the combo names created in OmniRoute. For example, if you created a combo named 'Auto-Claude-Haiku', add a model with that exact name in Claude Desktop.

Model discovery often fails; manual model entry is required

Enabling model discovery typically returns errors. Instead, manually add each model by copying the combo names from OmniRoute and entering them individually in Claude Desktop's model configuration.

Troubleshooting and Optimization

First API calls take longer due to initialization

The first request to any model (especially Nvidia) will have higher latency as the service initializes. Subsequent requests are faster. This is normal and expected.

Model sequence priority can cause failures

If a model combo fails, edit it and reorder the model sequence, moving more reliable models higher in priority. Switch from priority mode to round-robin if issues persist.

Logs in OmniRoute show real-time request details

Right-click the OmniRoute combo and open its logs in a new tab to see which model was called, response times, and any errors. This helps debug routing issues.

Use ChatGPT to extract model names for manual entry

When model discovery fails, copy all available models from OmniRoute into ChatGPT and ask it to list all model names. Use this list to manually add models to Claude Desktop.

Real-World Usage

Claude Desktop can now access project folders and files

With the gateway configured, Claude Desktop can analyze files in your local directories. Select a folder, and Claude will use whichever model combo you've configured to process documents and answer questions about them.

No subscription limits or token restrictions

By aggregating free providers, you bypass individual provider rate limits. If one provider's free tier is exhausted, another combo automatically handles the request.

Notable quotes

From now on, you won't have to pay a single rupee for Claude Desktop. — AI with FZ
Inside one Claude, all original Claude models are running. NVIDIA's free models are running. OpenAI's free models are running. — AI with FZ
It took me four days to debunk and develop it. If it works for you, please say in the comments. — AI with FZ

Action items

  • Install OmniRoute globally using: npm install -g omniroot
  • Launch OmniRoute and authenticate with Gmail to connect AntiGravity (Claude models)
  • Create a free Nvidia account and generate an API key, then add it to OmniRoute
  • Create model combos in OmniRoute for each provider (e.g., Auto-Claude-Haiku, Auto-Nvidia-Nemotron)
  • Uninstall Claude Desktop completely, then download and reinstall from claude.ai
  • Enable Developer Mode in Claude Desktop (Help > Troubleshooting > Enable Developer Mode)
  • Configure the gateway: paste OmniRoute endpoint as base URL and API key in Developer settings
  • Manually add each model combo name to Claude Desktop's model configuration
  • Test each model combo by sending a message and verifying the response in OmniRoute logs
  • Adjust model sequence or switch to round-robin mode if any combo fails consistently

More like this