Deploy a private local agent stack with Self-Hosted AI Stack
Stand up a local-first Docker Compose backend for agent work: LLM routing, chat, document parsing, embeddings, speech, and MCP tools.
npx skills add agentskillexchange/skills --skill deploy-private-local-agent-stack-with-self-hosted-ai-stack
Use this skill when an operator needs to provision a private, local-first AI backend that agents can call through OpenAI-compatible APIs and MCP tools. The workflow is to clone Self-Hosted AI Stack, choose the full or lightweight Compose profile, start the services, pull a local model, retrieve generated service keys, run the built-in health check, and connect agent clients such as goose, Cline, Claude-compatible tools, or other OpenAI-compatible clients through GatewayCrate and ToolUplink.
What this skill actually does
Invoke this instead of using a hosted product normally when the task requires local model execution, private document/RAG workflows, private speech or transcription services, or a reusable MCP tool layer on infrastructure the operator controls. The scope is deployment and verification of the private agent backend, not general use of every bundled application. Keep the operator focused on Docker Compose setup, model/key bootstrap, health checks, endpoint handoff, and the security boundary for local or HTTPS-exposed deployments.
Inputs and prerequisites: Docker, Docker Compose, Ollama/InferCrate, GatewayCrate/LiteLLM, ToolUplink/MCPHub, AnythingLLM.
Setup notes: git clone https://github.com/hwdsl2/self-hosted-ai-stack && cd self-hosted-ai-stack && docker compose up -d && docker exec ollama ollama_manage –pull llama3.2:3b && ./stack-check.sh
Source and verification boundary: use https://selfhostedaistack.com as the canonical reference before running the workflow; keep commands, API calls, CLI usage, and generated outputs reviewable against that upstream source.
Framework fit: publish this as a Multi-Framework workflow only when the operator can invoke the documented toolchain directly, rather than treating the upstream project as a generic product listing.