Tools overview

MiniTavern can use a Custom model link whenever the backend speaks OpenAI-compatible Chat Completions (or can map to it). Use the table to pick a tool.

Comparison

Tool	Difficulty	UI	Pull models	Roleplay fit	Typical Base URL
Ollama	★☆☆	Tray + CLI	ollama pull	Strong general chat	http://IP:11434/v1
LM Studio	★☆☆	Full GUI	In-app download	General	http://IP:1234/v1
KoboldCPP	★★☆	Light GUI/CLI	Bring your own GGUF	Common in RP communities	http://IP:5001/v1
llama.cpp server	★★★	CLI	Bring your own GGUF	Controllable, light	http://IP:8080/v1
LocalAI	★★☆	API / optional UI	YAML / images	Multi-backend	http://IP:8080/v1
Jan	★☆☆	Desktop app	In-app	General	See app “API” page
text-generation-webui	★★★	Web UI	Many extensions	Plugin ecosystem	After OpenAI extension

How to choose (one line)

1. Just get it working: Ollama or LM Studio → 30-minute quickstart
2. Care about RP presets / samplers: KoboldCPP
3. Have GGUF, want minimal deps: llama.cpp llama-server
4. Already on Docker / want a compatibility layer: LocalAI
5. Want an all-in-one desktop assistant: Jan
6. Already run oobabooga: enable the OpenAI-compatible extension, then connect MiniTavern

Shared requirements

Whatever you pick:

1. Listen on the LAN (0.0.0.0 or “allow LAN devices”), not only 127.0.0.1
2. Allow the port in the firewall
3. On the phone use the PC LAN IP, Base URL with /v1 (unless that tool’s docs say otherwise)
4. Model name must exactly match the id / name shown in the tool

Next: Recommended models · Configure in MiniTavern