A Wrapper Library for Koboldcpp

KoboldCpp for DeepSeek Harness - a tool plugin that lets the harness online model hand repetitive text and vision (OCR) labor to a local KoboldCpp (llama.cpp) server.

opencode plugin that auto-discovers models from OpenAI-compatible providers and enriches them with context window sizes, capability flags, and model metadata. Supports Atlas, Ollama, oMLX, vLLM, TGI, SGLang, LM Studio, KoboldCpp, llama.cpp, and LocalAI.

MCP server that gives online LLMs local hands and eyes: OpenAI-compatible local inference (KoboldCpp / Unsloth / llama.cpp / LM Studio / Ollama) for cheap text work and vision work (OCR / image analysis / comparison).

opencode plugin that auto-discovers models from OpenAI-compatible providers and enriches them with context window sizes, capability flags, and model metadata. Supports Actualyze, Ollama, oMLX, vLLM, TGI, SGLang, LM Studio, KoboldCpp, llama.cpp, and LocalA

An api to query local language models using different backends

Agent Smith: terminal client for language model agents

[](https://www.npmjs.com/package/infergui)

Connect the Pi coding agent to local & self-hosted LLMs — Ollama, LM Studio, llama.cpp, llama-swap, vLLM, and any OpenAI-compatible server. Auto-discovery, model switching, and zero-JSON in-TUI onboarding.

Terminal-native local AI assistant — offline, GPU-accelerated, agentic. Runs any GGUF model locally with full tool use.

MCP server for KoboldAI API integration

MCP server for KoboldAI API integration

Compress LLM models for Ollama in one command. Keeps your originals, adds a -cx variant. Hardware-aware, fully local, no account required.

A new web where AI generates every page.

Live local-inference telemetry (tok/s, TTFT, prefill) pinned to the OpenCode sidebar.

Live inference telemetry pinned to the OpenCode 2 sidebar.

Pi extension that auto-discovers local LLM servers on your LAN first, then Tailscale, and registers each as its own Pi provider with dynamic models + context windows.