Can my existing AI tools talk to a local model? Yes, if they let you change the server address and speak the OpenAI format.
Short answer. Yes, if the tool lets you set its server address (the base URL) and speaks the OpenAI chat format, because Ollama, LM Studio and vLLM all expose an OpenAI-compatible server. For Ollama the address is http://localhost:11434/v1/ and for LM Studio http://localhost:1234/v1; Ollama’s clients must send an API key but the server ignores it. Tools that are services run by a vendor, such as ChatGPT or Microsoft Copilot, are not built to be pointed at your own model, so those stay in the cloud. Before you rely on a switch, check that the server supports the features the tool needs, that it is reachable by the people who need it, and who is recorded when they use it.
01
The address of each local server
Runtime
Default address
Notes
Ollama
http://localhost:11434/v1/
API key required by clients but ignored. Supports chat completions, completions, models, embeddings and non-stateful responses. Images only as base64; tool choice and log probabilities are not supported
LM Studio
http://localhost:1234/v1
Endpoints for models, responses, chat completions, embeddings and completions. Existing OpenAI clients work by changing the base URL
vLLM
Started with vllm serve and a model name
An HTTP server compatible with OpenAI’s Completions, Chat Completions, Embeddings and other APIs
Open WebUI
A web app, not a model server
Connects to Ollama or OpenAI-compatible APIs, and adds accounts, groups and local document search
02
How to switch a tool over
Find the setting. It is usually called base URL, API endpoint, custom provider or OpenAI-compatible server.
Enter the local server’s address, and any text as the key if the client insists on one.
Choose a model by the name the local server uses; ask the server for its list at /v1/models, which Ollama and LM Studio both support.
Test with a harmless prompt, then with a real task.
Check the features you rely on: streaming, tool calling, image input and embeddings.
03
What may not work
Features the local server lacks, such as log probabilities or tool choice on Ollama, or image URLs where only base64 images are accepted.
Vendor-only features. Stateful responses, hosted search and file tools belong to the vendor’s service; Ollama supports non-stateful responses only.
Tools that fix their own address. A hosted service such as ChatGPT or Microsoft Copilot sends requests to the vendor’s own models; we found no documented setting that points either at your own server.
Speed and quality, which depend on the model and hardware you run, not on the tool.
04
Reaching the server from other computers
Ollama binds to 127.0.0.1 by default and is exposed to the network with the OLLAMA_HOST variable. Wiz recommends not exposing Ollama to the internet unless it sits behind authentication such as a reverse proxy. Because Ollama ignores the API key, a shared server needs something in front of it to identify users and keep a record of who asked what.
05
Where ANVE fits
ANVE’s models are reachable today through the runtime’s own OpenAI-compatible endpoint, Ollama or vLLM. The governed ANVE API, with a key per user or agent and every request in the audit trail, is not on ANVE yet. Zoho and other tools can connect through MCP connectors that an administrator approves, and ready-made Google Workspace and Zoho connectors are not on ANVE yet. ANVE is pre-release. The security page lists the published positions.
06
Questions people ask
Do I need an API key for a local server?
Ollama’s documentation says the key is required by the client libraries but ignored by the server, so any string works. For a shared server that also means no per-user control, which you must add.
Does an OpenAI-compatible API send my data to OpenAI?
No. It only means the request format matches. Where the data goes depends on the server you point the tool at.
Can Microsoft Copilot or ChatGPT use my local model?
We found no documented setting that points either at your own model server. Both are hosted services; Microsoft describes Copilot as coordinating hosted large language models with Microsoft Graph content.
Which model name do I enter?
The name the local server uses, such as the tag you pulled in Ollama. Ask the server for its list at /v1/models, which Ollama and LM Studio both support.
07
What is not real yet
No sealed ANVE unit exists; every image of the appliance on this site is a labelled render. The software runs today on a development machine. No customers yet, no certifications, no delivery date, and the final price is not announced.
Reservations are ₹999 in India or US$12 elsewhere, refundable in full until you confirm an order in writing (terms). Offices that cannot use cloud AI can apply to be one of ten design partners.
