How to Use DeepSeek, Claude & OpenRouter With Janitor AI
Janitor AI’s built-in options can be enough for casual conversations, but some users want to connect external AI services for access to different models, providers, or API configurations. Three names that commonly appear together are DeepSeek, Claude, and OpenRouter.
They are not the same type of service. DeepSeek is an AI company and model ecosystem. Claude is Anthropic’s family of AI models. OpenRouter is an API routing platform that provides access to models from multiple providers through a unified API.
This guide explains how these services can work with Janitor AI, how Proxy/API configurations work, what information you need, and how to troubleshoot common configuration errors.
Because Janitor AI’s interface and provider model catalogs can change, always verify current field names, endpoints, and model IDs before configuring an API.
For general Janitor AI performance problems rather than API configuration issues, see our guide to Janitor AI slow responses.

Quick Answer
You can use compatible external API routes with Janitor AI’s Proxy settings when the current Janitor AI configuration supports the route you’re trying to use.
A typical setup requires:
- A Janitor AI account with access to the relevant API/Proxy settings
- An account with the API provider or routing service
- An API key
- A compatible API endpoint
- A current model identifier
- Credits or billing access when required
The most important rule is:
The API key, endpoint, API format, and model ID must be compatible with the same API route.
For example, an OpenRouter API key should be used with an OpenRouter-compatible endpoint and an OpenRouter model ID. A direct DeepSeek configuration should instead use the credentials, endpoint, and model identifier supported by DeepSeek’s API.
Do not copy model IDs from old tutorials without checking them first. DeepSeek’s legacy deepseek-chat and deepseek-reasoner API names were retired in July 2026, so older configuration guides can now be misleading.
DeepSeek vs. Claude vs. OpenRouter
The first step is understanding what each service actually is.
| Option | What it is | API route | Model selection | Best suited for |
|---|---|---|---|---|
| DeepSeek | AI company and model ecosystem | Direct DeepSeek API or compatible routing service | DeepSeek models | Users who specifically want DeepSeek |
| Claude | Anthropic’s AI model family | Direct Anthropic API or compatible routing service | Claude models | Users who specifically want Claude |
| OpenRouter | Multi-provider API routing platform | Unified OpenAI-compatible API | Models from multiple providers | Users who want broad model choice |
OpenRouter is therefore not another model competing directly with DeepSeek or Claude. It acts as a routing layer through which users can access models from different providers.
Its current documentation provides an OpenAI-compatible API and a large, changing catalog of models and providers.
How External Models Connect to Janitor AI
There are three basic connection patterns.
1. Direct provider API
Janitor AI → DeepSeek API → DeepSeek model
or:
Janitor AI → Anthropic API → Claude model
The provider receives the request directly.
2. OpenRouter
Janitor AI → OpenRouter → Selected model/provider
You use an OpenRouter API key and select a model available through its current catalog.
3. Another compatible routing service
Other services may provide compatible API endpoints that can be used through Janitor AI’s Proxy configuration.
The important point is that Janitor AI’s Proxy setting does not automatically mean OpenRouter. The endpoint and credentials determine which service receives the request.
What You Need Before Connecting an API
Before configuring an external service, make sure you have:
Janitor AI access
You need access to the current API/Proxy configuration area. The exact location and field names can change as the interface is updated.
A provider account
Depending on your route, this could be DeepSeek, Anthropic, OpenRouter, or another compatible service.
An API key
An API key authenticates your requests. Treat it like a password, especially because API usage can potentially generate charges.
A compatible endpoint
The endpoint is the URL where the API accepts requests. Different services and API formats can use different URLs.
A current model ID
The model ID tells the API which model should process the request. Always copy it from the current provider or routing-service documentation.
Credits or billing access
External API usage is separate from ordinary Janitor AI access. Depending on the service and model, you may need an account balance, billing setup, or other access.
API Terms You Should Know
Understanding a few terms makes the setup much easier.
API: A structured way for applications to communicate with another service.
API key: A credential used to authenticate API requests.
Endpoint: The URL where the API receives requests.
Model ID: The exact identifier used to select a model.
Proxy: In this context, a Janitor AI configuration that routes requests to an external compatible API endpoint. It is not the same as a VPN or network proxy.
OpenAI-compatible API: An API that follows an OpenAI-style request format. It does not mean the underlying model was created by OpenAI.
How Janitor AI Proxy Settings Work
Current third-party Janitor AI integration guides commonly describe a chat-level API Settings area with a Proxy configuration containing fields such as:
- Configuration name
- Model name
- Proxy/API URL
- API key
The exact interface can change, so treat current Janitor AI instructions as the source of truth for navigation.
The key compatibility relationship is:
API key
+
Endpoint
+
API format
+
Model ID
=
Compatible API route
It is not necessary for every component to belong to the same company.
For example:
Janitor AI
↓
OpenRouter API key
↓
OpenRouter endpoint
↓
OpenRouter model ID
↓
Claude model
The underlying model can be Claude even though the API key and endpoint belong to OpenRouter.
OpenAI-Compatible vs. Anthropic-Compatible APIs
This distinction is important because a provider can offer more than one API format.
DeepSeek currently documents both an OpenAI-compatible API and an Anthropic-compatible API. Its current documentation lists:
OpenAI-compatible base URL:
https://api.deepseek.com
Anthropic-compatible base URL:
https://api.deepseek.com/anthropic
The exact endpoint required by Janitor AI may differ from the provider’s documented base URL. Some integrations require a complete endpoint rather than only a base URL.
The important lesson is:
A provider having an API does not mean every application can use every API format that provider offers.
Use the API format and endpoint supported by the Janitor AI configuration you’re setting up.
How to Use DeepSeek With Janitor AI
DeepSeek currently documents V4 models including:
deepseek-v4-flashdeepseek-v4-prodeepseek-v4-flash-vision-exp
Model availability can change, so verify the current model list before configuring Janitor AI.
Direct DeepSeek setup
Step 1: Create your DeepSeek API account
Use DeepSeek’s current API platform and complete the required account setup.
Step 2: Generate an API key
Create an API key from the provider’s current API dashboard and keep it private.
Step 3: Check the endpoint
DeepSeek currently documents:
https://api.deepseek.com
for its OpenAI-compatible API and:
https://api.deepseek.com/anthropic
for its Anthropic-compatible interface.
However, don’t automatically paste a base URL into Janitor AI if its Proxy field expects a complete endpoint.
Follow the endpoint format required by the current Janitor AI configuration and the API format you’re using.
Step 4: Select a current model
Use a model ID currently listed by DeepSeek, such as:
deepseek-v4-flash
or:
deepseek-v4-pro
Do not assume an older model name is still supported.
Step 5: Add the configuration
Open a Janitor AI chat and locate the current API Settings/Proxy area.
Enter the required:
- Configuration name
- Model ID
- Compatible API/proxy URL
- API key
Step 6: Save and test
Save the configuration and send a short test message.
For example:
Reply with one short sentence confirming that the connection works.
If the request fails, check the API configuration before changing prompts, temperature, character settings, or other generation options.
Why deepseek-chat May Not Work
Older Janitor AI tutorials often use:
deepseek-chat
Do not copy this model ID blindly.
DeepSeek retired its legacy deepseek-chat and deepseek-reasoner API names in July 2026. Current DeepSeek documentation lists its newer V4 model identifiers instead.
This is one of the clearest examples of why API instructions should be checked against current provider documentation rather than copied from an old tutorial.
Direct DeepSeek vs. DeepSeek Through OpenRouter
| Factor | Direct DeepSeek | DeepSeek through OpenRouter |
|---|---|---|
| API account | DeepSeek | OpenRouter |
| API key | DeepSeek key | OpenRouter key |
| Endpoint | DeepSeek API route | OpenRouter API route |
| Model selection | DeepSeek models available on the route | Models available through OpenRouter |
| Billing | DeepSeek account | OpenRouter account |
| Routing | Direct provider connection | OpenRouter adds a routing layer |
| Main advantage | Direct provider relationship | Easier access to multiple providers |
| Main drawback | Limited to the provider’s ecosystem | Additional service involved |
Neither option is automatically faster, cheaper, or more reliable. Those factors can vary by model, provider, routing, account limits, and current service conditions.
How to Use Claude With Janitor AI
Claude is Anthropic’s family of AI models.
It is important to distinguish Claude’s consumer service from Anthropic API access. Having access to Claude through a consumer subscription does not automatically mean you have API credits for an external Janitor AI configuration.
For Janitor AI, determine which Claude API route the current configuration supports.
Possible routes can include:
- A direct Anthropic-compatible route, if supported by the current Janitor AI configuration.
- OpenRouter, where relevant Claude models are available through OpenRouter’s API.
General Claude setup
- Create the required API account for your chosen route.
- Generate an API key.
- Check the route’s current API documentation.
- Copy the current Claude model identifier.
- Open Janitor AI’s API Settings/Proxy area.
- Enter the endpoint required by that API route.
- Enter the matching API key.
- Enter the current model ID.
- Save the configuration.
- Send a short test message.
Do not assume a Claude model ID from OpenRouter is interchangeable with an Anthropic-native API model ID. The underlying model family may be the same while the API route and identifier are different.
Anthropic also updates and retires model versions, so use its current model documentation when configuring Claude.
How to Use OpenRouter With Janitor AI
OpenRouter is useful when you want to access models from multiple providers through one API platform.
Its current documentation describes an OpenAI-compatible API and a large catalog of models and providers.
Step 1: Create an OpenRouter account
Create an account and open the API key section.
Step 2: Generate an API key
Create a key and store it securely.
Do not publish it in screenshots, forums, Discord channels, GitHub repositories, or support comments.
Step 3: Choose a model
Use OpenRouter’s current model catalog to find the model you want.
OpenRouter uses provider/model-style identifiers, so copy the exact identifier shown for the model rather than relying on an old tutorial.
Step 4: Configure Janitor AI
In Janitor AI:
- Open a chat.
- Open API Settings.
- Select Proxy.
- Add a configuration.
- Enter a configuration name.
- Enter the current OpenRouter model ID.
- Enter the OpenRouter-compatible URL required by Janitor AI.
- Enter your OpenRouter API key.
- Save the configuration.
- Refresh if necessary and test the connection.
OpenRouter currently documents:
https://openrouter.ai/api/v1
as its OpenAI-compatible API base URL and:
https://openrouter.ai/api/v1/chat/completions
for chat-completion requests.
Because Janitor AI’s Proxy field may expect a complete endpoint rather than only a base URL, follow the current Janitor AI integration requirements when entering the URL.
Which Route Should You Use?
There is no universal winner.
| Your priority | Route to consider | Why |
|---|---|---|
| Use DeepSeek specifically | Direct DeepSeek | Direct connection to DeepSeek |
| Use Claude specifically | Direct Anthropic, if supported | Direct provider route |
| Try many providers | OpenRouter | Multiple models through one API platform |
| Compare models | OpenRouter | Centralized model selection |
| Minimize routing layers | Direct provider | Fewer services involved |
| Experiment with different models | OpenRouter | Large changing model catalog |
If you only want one provider, a direct connection may be simpler. If you frequently switch between different model families, OpenRouter can be more convenient.
DeepSeek vs. Claude: Which Is Better?
There is no permanent answer because model capabilities, versions, pricing, and availability change.
Instead, compare the specific models you intend to use based on:
- Availability: Is the model supported by your chosen API route?
- API access: Does your account have access to the required API?
- Cost: What does the current provider charge?
- Context: Does the model provide the context capabilities you need?
- Writing or roleplay quality: Test the actual model rather than relying on old rankings.
- Setup complexity: Direct APIs generally involve fewer services, while OpenRouter can simplify switching between providers.
The best choice is usually the model that performs well for your specific use case at an acceptable cost and availability.
When Should You Use OpenRouter?
OpenRouter makes particular sense when you want to:
- Compare models from different providers.
- Switch between providers.
- Use one API platform for multiple model families.
- Explore a large and frequently changing model catalog.
The tradeoff is that OpenRouter adds another service between Janitor AI and the underlying provider:
Janitor AI
↓
OpenRouter
↓
Selected provider/model
If something fails, you may need to determine whether the problem is with Janitor AI, OpenRouter, the selected model, the underlying provider, your account, or your network.
For a single-provider setup, a direct API route may therefore be easier to troubleshoot.
Does Using DeepSeek, Claude, or OpenRouter Cost Money?
External API costs and Janitor AI access are separate considerations.
When you connect Janitor AI using your own external API credentials, usage may be billed according to the provider or routing platform’s current pricing.
DeepSeek currently documents API pricing based on input and output tokens.
OpenRouter displays pricing on individual model listings, and prices can vary significantly between models and providers.
Some OpenRouter models may have free variants or other no-cost conditions, but these offerings can change.
Therefore:
Never assume a model described as “free” in an old Janitor AI tutorial is still free.
Check the current pricing page for the specific model and route you plan to use.
How to Keep Your API Key Safe
An API key can provide account access and potentially generate usage charges.
Treat it like a password.
Never:
- Publish your API key.
- Put it in a screenshot.
- Upload it to GitHub.
- Post it in Reddit or Discord.
- Paste it into an unknown troubleshooting website.
- Give it to an unofficial setup service.
Do:
- Store the key privately.
- Monitor API usage.
- Review billing activity.
- Use spending controls where available.
- Revoke or rotate an exposed key immediately.
You never need to publish your real API key to get help diagnosing a Janitor AI configuration.
Janitor AI API Errors
| Error or symptom | Often points to | What to check |
|---|---|---|
| 401 Unauthorized | Authentication | API key, account, endpoint |
| 403 Forbidden | Access or permissions | Account status, model access, restrictions |
| 404 Not Found | URL or model | Endpoint path and model ID |
| 429 Too Many Requests | Rate, concurrency, or usage limit | Provider limits and account usage |
| Invalid model | Model identifier | Current model ID |
| Network error | Connection or service | Network, endpoint, provider |
| Timeout | Request did not finish | Network, routing, provider, Janitor AI |
| Insufficient balance | Billing | Provider or routing account |
These are diagnostic starting points rather than absolute rules.
For example, a 429 response can indicate rate limits, concurrency limits, quotas, or other usage restrictions. DeepSeek’s current API documentation specifically documents concurrency limits.
Troubleshooting Workflow
If your configuration fails, check the following in order.
1. Confirm the API route
Ask:
Who issued the API key?
Was it DeepSeek, Anthropic, OpenRouter, or another service?
2. Confirm the endpoint
Make sure the endpoint belongs to the same compatible API route.
For example, don’t combine an OpenRouter key with a direct DeepSeek endpoint.
3. Confirm the model ID
Copy the current model identifier directly from the provider or routing-service catalog.
4. Check the API key
Look for:
- Extra spaces
- Incomplete copying
- Wrong provider
- Revoked key
- Incorrect account
- Restricted credentials
5. Check the provider account
Review:
- Balance
- Billing status
- Usage
- Rate limits
- Account restrictions
- Model availability
6. Test a clean chat
Start a new conversation and send a short message. This removes long conversation history and character-specific settings from the initial test.
7. Test another model
If the route supports multiple models, test another currently available model.
If one model fails while another works, the problem may be model-specific.
8. Check Janitor AI
If the API key, endpoint, model, and provider account all appear correct, investigate Janitor AI itself.
For broader performance problems, see our guide to troubleshooting Janitor AI performance.
The 5-Minute Janitor AI API Check
Before spending an hour changing settings, run this quick test:
Minute 1 — Provider: Confirm where your API key came from.
Minute 2 — Endpoint: Confirm the URL matches that API route.
Minute 3 — Model: Copy the current model ID from the live provider catalog.
Minute 4 — Credentials: Confirm the key is active and the account has the required access or balance.
Minute 5 — Test: Open a clean chat and send one short message.
Think of the troubleshooting order as:
Credential → Endpoint → Model → Account → Service
This is more effective than randomly changing generation settings.
Common Janitor AI API Mistakes
Using the wrong API key
An OpenRouter key cannot simply replace a direct DeepSeek or Anthropic credential.
Mixing endpoints
The key and endpoint must belong to a compatible API route.
Copying an outdated model ID
Model identifiers can be retired or replaced.
Using a friendly model name
The API normally expects the exact model identifier shown by the provider.
Using the wrong URL format
Some integrations require a complete chat-completions endpoint rather than only a base URL.
Assuming every OpenRouter model is free
Pricing and free availability can change.
Assuming every model works through every route
A model available through OpenRouter is not automatically available through a direct provider endpoint.
Sharing an API key
Never share your real key publicly.
Changing multiple settings at once
Change one variable at a time so you can identify the cause of a problem.
Assuming every error means Janitor AI is down
The issue may be the provider, model, account, endpoint, routing service, or network.
Using an unknown proxy
Be cautious with proxy URLs copied from old forums or social media posts. You should understand who operates the service and how requests are handled before using it.
Why Old Janitor AI API Tutorials Stop Working
API tutorials can become outdated even when their original instructions were correct.
Things that can change include:
- Model IDs
- API endpoints
- Pricing
- Free-model availability
- Janitor AI interface fields
- Account requirements
- API compatibility
- Provider model availability
DeepSeek is a clear example. Its older deepseek-chat and deepseek-reasoner identifiers were retired in July 2026, while its current documentation lists newer V4 model identifiers.
OpenRouter also maintains a frequently changing model catalog.
Before following an old tutorial, verify:
- Current endpoint
- Current model ID
- Current provider documentation
If those three match, the rest of the configuration is much easier to troubleshoot.
Frequently Asked Questions
Can you use DeepSeek with Janitor AI?
Yes, when the current Janitor AI Proxy configuration supports the required API route. DeepSeek can be accessed directly through its API or through a compatible routing platform such as OpenRouter.
Can you use Claude with Janitor AI?
Claude can be accessed through compatible API routes, including OpenRouter when the relevant Claude model is available. Claude consumer access should not be confused with separate API access.
Can you use OpenRouter with Janitor AI?
Yes, when the current Janitor AI configuration supports the compatible OpenRouter route. OpenRouter provides an OpenAI-compatible API and access to models from multiple providers.
How do I add an API key to Janitor AI?
Open a Janitor AI chat, locate the current API Settings area, select Proxy, and create or edit a configuration. Enter the compatible model ID, API URL, and API key, then save and test the configuration.
Is OpenRouter free?
OpenRouter supports models with different pricing arrangements. Some models may have free variants, while others are paid. Check the current listing for the specific model you want to use.
Is DeepSeek free with Janitor AI?
Not automatically. Direct DeepSeek API usage follows DeepSeek’s current API pricing, while a route through OpenRouter follows OpenRouter’s pricing for the selected model.
Why is my Janitor AI API key not working?
First check whether the API key, endpoint, and model ID belong to a compatible API route. An incorrect provider, outdated model ID, revoked key, account restriction, or incorrect endpoint can cause authentication or configuration failures.
What does a 401 error mean in Janitor AI?
A 401 response generally points to an authentication problem. Check that the API key is correct and active and that it is being sent to a compatible endpoint.
What does a 404 error mean in Janitor AI?
A 404 often points to an incorrect endpoint or model identifier. Check the complete URL and copy the current model ID from the relevant provider or routing service.
What does a 429 error mean in Janitor AI?
A 429 response generally indicates a rate, concurrency, quota, or usage restriction. Check the provider’s current limits and account usage.
Why does my model ID not work?
Model identifiers can change when providers release, deprecate, or retire models. Copy the current identifier directly from the service you’re using.
Can I switch AI models after setting up Janitor AI?
Yes, when the new model is supported by your API route and configuration. With OpenRouter, for example, different models can be selected from its current catalog.
Conclusion
DeepSeek, Claude, and OpenRouter play different roles in the AI ecosystem. DeepSeek and Claude represent model ecosystems, while OpenRouter is a routing platform that provides access to models from multiple providers.
When connecting an external service to Janitor AI, the most important thing is understanding the relationship between the API key, endpoint, API format, and model identifier.
If a configuration stops working, verify those components against the provider’s current documentation before changing other settings. Model catalogs, pricing, endpoints, and interfaces can change over time.
Most importantly, keep your API key private. Never publish it or give it to an unofficial troubleshooting service.
For a broader explanation of Janitor AI itself, see our complete Janitor AI guide.
Official Documentation
For current configuration details, always verify the provider’s official documentation before entering API credentials, endpoints, or model IDs.
- DeepSeek API: Current models, endpoints, API formats, pricing, and usage limits.
- OpenRouter: Current model catalog, API endpoints, routing options, and documentation.
- Anthropic/Claude: Current Claude API documentation and model lifecycle information.
