ipsupport

IPSupport LLC / Open source

Local AI, useful integrations, and infrastructure you control.

Explore our public projects: from running models on your Mac to connecting assistants with your work and sharing compute across a Kubernetes cluster.

Local AI and agents

Run models. Build with them.

Desktop tools and services for working with local and cloud models.

macOS · Apple Silicon

LLMTray

A native Mac app for local AI: chat with tools, browse and download models, generate and edit images, and create music. Serves OpenAI- and Anthropic-compatible APIs for your apps and coding agents.

Go · Terminal agent

ipsupport-code

A lightweight coding agent for local and cloud models. Plan and auto modes, workspace permissions, a built-in risk check, MCP, skills, sub-agents, and lessons saved between runs.

Go · Self-hosted gateway

AirLLM

Route agents to language models through one gateway, with model policies, usage limits, provider fallback, token and cost metering, prompt scanning for secrets and PII, and an admin console.

MCP connectors

Connect assistants to your daily work.

Tools for any MCP client on macOS, built with JavaScript for Automation and no extra runtime dependencies.

macOS · MCP

Apple Mail, Calendar & Notes

Read, search and organize mail, prepare replies, manage calendar events, and keep notes through the apps on your Mac. Includes read-only mode and previews for supported write operations.

macOS · MCP

Seafile MCP

Browse, search and organize Seafile libraries, transfer files, restore versions, and manage share links. Browser sign-in stores the token in macOS Keychain; read-only mode is available.

Infrastructure and distribution

Put your own hardware to work.

Kubernetes tools for shared GPUs and on-demand compute, plus packages for installing our apps.

Go · Kubernetes DRA

dra-driver-dri

Share GPUs between Kubernetes pods by requested VRAM. Exposes Linux DRM render nodes through CDI for Vulkan inference, including older AMD hardware without ROCm. VRAM requests guide scheduling; they are not hard memory limits.

Go · Kubernetes autoscaling

plugpool

A cluster-autoscaler provider for bare-metal machines powered by ESPHome smart plugs. Boots nodes when capacity is needed and shuts them down before cutting power, with support for scaling to zero.

Homebrew · macOS and Linux

Homebrew tap

Install LLMTray or LLMTray Full on Mac, and ipsupport-code (also available as ipco) on macOS and Linux. Formulae and casks track stable product releases.

Public forks

Our copies of upstream projects.

These repositories are forks. Follow each repository for its upstream attribution and the changes in our copy.

Fork · MLX

mlx-lm

Text generation, quantization and fine-tuning with MLX on Apple Silicon. LLMTray runs a pinned version of this fork.

Fork · MLX

mlx-audio

Speech synthesis, transcription, speech-to-speech and music generation with Apple’s MLX framework.

Fork · MLX

mflux

Native MLX implementations of generative image and video models for running locally on supported hardware.

Fork · Terraform

terraform-null-label

A reusable context module for consistent resource names, labels and tags. Pass context between modules across cloud and on-premises infrastructure.