Ollama Clients
/v1/chat/completions. Clients that speak Ollama's native /api/chat or /api/tags will not work against Prism; clients that support an OpenAI-compatible base URL will.Last updated Reviewed against Prism v0.3.26
What Prism actually serves
| Route | Available on Prism? |
|---|---|
| /v1/chat/completions | Yes — OpenAI Chat Completions. |
| /v1/responses | Yes — OpenAI Responses. |
| /v1/messages | Yes — Anthropic Messages, for Anthropic-protocol clients. |
| /v1/models | Yes — your routed catalog, no auth needed. |
| /api/chat | No. This route does not exist in Prism. |
| /api/generate | No. |
| /api/tags | No. Ollama's model listing is not served. |
| /api/show | No. |
/api/chat examples in earlier Prism documentation were incorrect.Prism does speak Ollama upstream
The reverse direction is supported: a custom provider can point at a real local Ollama server over its native API, using Ollama's own /api/chat format as the upstream. That is an outbound setting on a provider, not an inbound route on Prism. Configure it in Provider Setup.
Tools that work today
Any client with an OpenAI-compatible base URL works. Point it at http://127.0.0.1:11434/v1 with prism as the key.
| Client | What to set |
|---|---|
| Continue.dev | An OpenAI-compatible model with apiBase http://127.0.0.1:11434/v1. |
| Open WebUI | An OpenAI connection with the base URL above, rather than the Ollama connection. |
| LibreChat | An openAI-style endpoint in its config with the Prism base URL and key. |
| Cline, Roo and similar VS Code extensions | An OpenAI-compatible provider with the same base URL. |
| Any SDK-based script | Set OPENAI_BASE_URL and OPENAI_API_KEY. |
# Confirm Prism is answering before configuring a client curl http://127.0.0.1:11434/v1/models
Running Prism alongside a local Ollama
Both default to port 11434, so only one can hold it. If Ollama runs as a background service, move Prism to another port with PRISM_PORT.
# PowerShell $env:PRISM_PORT = "11435" # Prism now serves http://127.0.0.1:11435/v1 # bash export PRISM_PORT=11435
Changing the port also moves the base URL written into every agent integration, so re-sync your agents from the Agents tab after changing it.
http://127.0.0.1:11434 for local models. Your agents then reach local and cloud models through one base URL.Model names
Ollama-oriented tools often default to names like llama3 or codellama. Prism serves a request for an unknown model from your default model, or you can alias the name explicitly:
{
"aliases": {
"llama3": "glm-5.1:cloud",
"codellama": "deepseek-v4-flash:cloud"
}
}