The llama-cpp provider either installs a pinned, integrity-verified llama-server under OpenClaw's localService supervisor or connects to a server that you already operate. Both choices use llama-cpp/<model> references and OpenClaw's normal OpenAI-compatible chat transport. Local embeddings require the managed choice.

Install

openclaw plugins install @openclaw/llama-cpp-provider

Restart the Gateway after installing or updating the plugin. Interactive setup shows Managed local server and Existing llama-server under one Local llama.cpp group.

Configure managed text inference

After explicit consent, OpenClaw installs the matching server build and downloads Gemma 4 E4B IT Q4_K_M (approximately 5.0 GB) plus EmbeddingGemma (approximately 0.3 GB). The default chat download is offered only on machines with at least 16 GiB of RAM.

Custom GGUF models remain supported through params.modelPath. Rerun llama.cpp setup after changing the model so OpenClaw can verify the file and regenerate the managed router preset.

See the llama.cpp provider guide for platform requirements, custom GGUF configuration, diagnostics, and repair.

Connect to an existing server

Choose Existing llama-server during setup and enter the endpoint and optional API key. OpenClaw passively discovers single-model and router catalogs. It never installs, starts, stops, or reconfigures the external process.

See the llama.cpp provider guide for authentication, router behavior, manual configuration, and troubleshooting.

Configure embeddings

Set memory.search.provider to local. The plugin preserves the historical local embedding provider and index identity while serving requests through the managed server's /v1/embeddings endpoint.

Package

Plugin id: llama-cpp
Provider id: llama-cpp
Package: @openclaw/llama-cpp-provider
Minimum OpenClaw host: 2026.6.2