Local AI gateway
When your models, MCP services, or search provider changes, update the gateway once instead of every Agent.
Previously, changing large models meant updating every AI tool you were using one by one, whether you were changing the API key or switching provider. Now, you only need to make the change in OmniProxy.
See how it worksSeeing the details of each exchange between an AI tool and a model is an effective way to learn how AI works. When you give an AI tool a task, you usually see only part of the model’s response. Open OmniProxy to see how the tool and model communicate.
See how it worksClassify models and filter the models a user can use based on their current device, helping users quickly find the local model that suits them among tens of thousands of open-source models.
See how it worksSet up text redaction rules for each external source. Once enabled, sensitive text such as passwords and phone numbers will be replaced with redaction markers before being sent to the corresponding source, helping to reduce the risk of privacy and confidential information leaking.
See how it worksConnect your chosen web search service to obtain web page summaries and original links on demand, giving the large model more references.
See how it worksSet up a professional glossary for translation models to help translations use the industry terms you need.
See how it worksChat, image generation, read-aloud and video generation each have a fixed model entry point. When you want to try a new model, change the corresponding selection in OmniProxy and the AI tools will continue using the original entry point, without the hassle of reconfiguring them.
See how it worksYou don't have to hunt for options or look up commands yourself every time you want to check status or adjust settings. Let a command-capable AI tool do it for you through the OmniProxy command line, hand repetitive configuration work to AI, and free up time for your own work.
See how it worksOmniProxy.ai connects AI tools, models, and services through one gateway, bringing keys, model mapping, call logs, search, translation, and redaction together.
When you change model provider, there is no need to modify each AI tool's configuration one by one, nor to store model keys separately across different tools. Connection settings are managed centrally in OmniProxy, and keys are kept in the macOS system Keychain.
Read the chat history between AI tools and models to understand in depth how AI tools interpret tasks, organise requests and use the model's responses. This helps you use AI more effectively and provides clues for troubleshooting the issues you are currently facing.
Open-source models are organised by category and filtered for what your current device supports, helping you choose faster. You can specify multiple model folders and directly use models that other tools have already downloaded, reducing duplicate downloads. You can also download models to an external drive or network drive to save local storage space.
Replace sensitive information identified by rules in the text, such as phone numbers, bank accounts and passwords, with xxxxxx so that this content is not sent to remote services with the request.
Connect a web search service so that AI can obtain new information beyond its existing knowledge. Web page summaries and original links are also provided, making it easy for you to view sources and check the evidence behind answers.
Keep product names, industry terms and common expressions consistent, reducing the frustration of seeing the same word translated differently across paragraphs. Build a glossary from translations you approve and reuse it in supported translation workflows, reducing the repeated work of correcting terminology in each document.
Set up fixed model entry points for purposes such as chat, text-to-image, read-aloud and video generation. AI tools only need to connect to the relevant entry point. If you later want to change the model for that purpose, simply switch it in OmniProxy without modifying the AI tool's calling configuration.
Hand over the command-line-supported queries and settings to a command-capable AI tool, automate the various OmniProxy settings, and greatly save configuration time.
OmniProxy does not install or manage these third-party tools. Use each tool's official installer, then follow its connection guide.
Want to use local and cloud models in the same chat window but are concerned that deployment and connection setup will be too complicated? Open WebUI provides a familiar chat interface. Once installed, you can start using it by entering the connection details step by step.
Read installation and setupWant an AI assistant that can read files, write content, or handle repetitive tasks without being limited to a web browser? goose offers both desktop and terminal versions. Once installed and configured with a model, it is ready to use on your own computer.
Read installation and setupWant to install an AI assistant on your own computer but are concerned about setting up the environment or switching models? OpenClaw provides a desktop app and installation scripts, allowing you to use it from a messaging app or terminal after installation.
Read installation and setup
