Connect your AI Start for free, or bring your own AI
Start for free
- You will need
- CogineWork installed. No sign-up and no key
Open CogineWork
Cogine Free is built in and available the first time you open the app
Choose Cogine Free
Select it in the model picker. The free model stays free and does not expire
Start a task
Try a small, focused task first, then bring in your own material
A first task to try
Turn this document into three key points and mark anything I still need to confirm
When you want a stronger model later
Free models are a good way in. When you want a stronger model or more choice, connect your own AI using any method below. Gemini, OpenRouter, GitHub Models, OpenCode Zen, Kilo Code and Zhipu BigModel also offer free models you can use right away with your own account through an API key.
Connect Codex in CogineWork
- You will need
- A ChatGPT account with the relevant Codex access and a working internet connection
Choose the Codex connection
Find the OpenAI Codex connection in model settings. It is separate from OpenAI connections that use an API key
Authorize your account
Follow the browser authorization or device-code flow offered by the client. Use a ChatGPT account with the relevant Codex access
Choose a model and try a task
After authorization, choose an available model in the client. Check the connection with a focused task before bringing in your own material
A first task to try
Compare these two public sources. List what they agree on, where they differ and what needs more verification. Keep the sources.
If the connection does not complete
Check the account’s Codex access, your network connection and the authorization status in the client. Do not use a standard OpenAI API key in the Codex account flow. For API-key access, choose the corresponding API connection.
Connect OpenCode Go, OpenRouter and 20+ leading AI providers with an API key
- You will need
- An API key for the selected service and the usage allowed by that service plan
Choose the matching connection
Select OpenCode Go, OpenRouter, DeepSeek or another provider in the client. Go, Zen and each provider, region and plan have their own connections
Configure the key for that service
Enter the key for the selected service. Availability and charges follow your service plan
Choose a model and try a task
Available models and capabilities vary by connection. Confirm the connection with a small task first
If the connection does not complete
Model use follows the billing rules of your connected account or provider. Connecting a subscription does not move its allowance to another provider.
Connect a local model from Ollama, LM Studio or Hugging Face
- You will need
- A model suited to your hardware. Ollama and LM Studio must be running on this computer first
Choose a local source
Connect a running Ollama or LM Studio, or download a model that fits your hardware and task from Hugging Face inside CogineWork
Choose the local connection in the client
Connect to the local service or the downloaded model using the connection supported by the client
Choose a model and try a task
Speed and capability depend on your hardware and runtime. Confirm the connection with a small task first
If the connection does not complete
Local-first does not always mean no data leaves your computer. Cloud models and external tools may receive the content needed for a task; the data path depends on your chosen models, tools and configuration.
Steps reflect the current client version; connection names and authorization flows may change between releases