Mesh-LLMOfficial·28 Agent Skills Included

mesh-llm

Pool GPUs across machines into one shared AI model API

Pools GPUs and memory across multiple machines and exposes them as one OpenAI-compatible API for running large language models. Eliminates the need for a single expensive GPU server by splitting big models across ordinary machines with encrypted peer connections. Includes ready-made skills for deploying nodes, joining meshes, connecting coding agents, benchmarking, and validating releases.

3,350|405

↗

npx skills add Mesh-LLM/mesh-llm --all -g -y

Available:

Gives the AI agent a map of the repository, the mandatory build commands, and the strict rules for building, testing, and patching the runtime so it makes changes safely.

All Skills in This Repository (28)

Pure Emerald Level Indicators

📦 In Repo

Mesh-LLM

llama-patch-changes

Manage llama.cpp patch queues and upstream pins for Mesh-LLM builds.

metrics-server

Ingest OTLP telemetry into DuckDB and export benchmark reports via HTTP API.

skippy-family-certification

Certifies GGUF model families for staged splits via parity and topology checks.

skippy-server

Run the skippy-server backend runtime for binary stage transport and OpenAI integration.

telemetry-privacy-review

Audits OTLP metrics and telemetry attributes for privacy risks before changes.

skippy-metrics

Standardize telemetry attributes and benchmark metrics across Skippy components.

skippy-correctness

Validate staged Skippy execution parity against full-model results across deployments.

skippy-spec-bench

Benchmark target and draft GGUF pairs to verify decoding compatibility.

hf-layer-package-jobs

Automate planning, validation, and submission of Hugging Face GGUF layer-package jobs.

llama-stage-patch-changes

Regenerate the llama-stage.cpp patch queue and update the upstream SHA pin.

skippy-prompt

Orchestrates setup and lifecycle of prompt-owned staged workflows across multiple hosts.

kv-tool-loop-stability

Certify KV/cache stability of mesh-llm OpenAI-compatible endpoints under repeated tool-loop activity.

Official

Advanced

Frequently Asked Questions

FAQPage Schema

How to install Mesh LLM?▼

Run `npx skills add Mesh-LLM/mesh-llm --all -g -y` in your terminal to install all skills in this suite globally.

What does Mesh LLM actually do?▼

It connects GPUs across several computers into one mesh and serves large AI models through a single OpenAI-compatible API, so no single machine needs to hold the whole model.

Can I run models too big for one GPU?▼

Yes. Skippy split serving divides a model's layers across multiple nodes, and the mesh routes requests automatically with encrypted connections.

Does Mesh LLM work with coding agents like Claude Code?▼

Yes. Built-in launchers connect Goose, Claude Code, OpenCode, and Pi directly to your local mesh with one command, no manual configuration needed.

Do I need to download models manually?▼

No. Pass a model reference to the serve command and Mesh LLM downloads and loads the model itself on first use.