Deployment Recipes
Copy-paste setups that actually work. No marketing, no missing prerequisites, no "it should work on your machine." Just the commands, the expected output, and the fixes when something goes wrong.
44 recipes
Self-Host Agenta with Docker Compose (Agent Studio + Runner)
Agenta v2.0's self-hosted deployment runs the studio, API, agent runner, and datastores on your own infrastructure via Docker Compose — with optional Daytona cloud sandboxes for multi-user isolation.
Deploy OpenHands 1.0 in Docker (Autonomous Coding Agent)
OpenHands reached its 1.0 release with production-ready Docker sandboxing, built-in security policies, resource limits, and a plugin system — autonomously completing ~68% of SWE-bench Verified tasks. Here's the self-hosted Docker deployment recipe.
Deploy ComfyUI with Docker for Local AI Image Generation
Stand up ComfyUI in a Docker container on your own GPU for private, unlimited AI image generation — no API bills, no per-image fees, no prompts leaving your machine.
Deploy a Multi-Container Agent Stack with Docker Compose for Agents
Use Docker's Compose for Agents specification to define and run a complete self-hosted agent stack — runtime, LLM gateway, vector store, and memory — in a single declarative file.
Deploy Google Agent Executor (AX) on Kubernetes
Stand up Google's open-source distributed agent runtime on Kubernetes with Agent Substrate for suspendable, resumable, fault-tolerant agent execution at scale.
Deploy Qwen3.8-Flash-Next with vLLM
Run Alibaba's experimental Qwen4-architecture model locally with vLLM — 125B total / 6B active, n-gram embeddings, Qwen Sparse Attention. Open weights, OpenAI-compatible API.
Deploy IBM Granite 4.2 with SGLang
Serve the 30B reasoning model with switchable thinking mode using SGLang on a single H200 or B200 — with Docker, thinking-mode examples, and benchmark validation.
Deploy Trinity Agent Orchestration Platform
Self-host a 24/7 agent orchestration platform with cron scheduling, multi-agent delegation, fleet observability, and 116 MCP tools — open source under Apache 2.0.
Deploy NVIDIA Magpie TTS Multilingual Locally
Run NVIDIA's 357M open-weight text-to-speech model on your own GPU for low-latency multilingual voice agents — 12 languages, 32ms time-to-first-audio on B200.
Deploy Meta Muse Glimmer 30B Locally with Ollama
Run Meta's open-weight 30B agentic multimodal model on a single consumer GPU for local agent development — no API keys, no cloud.
Deploy an Agent Sandbox on Cloudflare Computer
Run AI agents in Cloudflare Durable Object workspaces with V8 isolates for lightweight tasks and full Linux containers for heavy execution — one filesystem, two isolation tiers.
Deploy vLLM with Flat Model Architecture
Set up vLLM's new Flat Model abstraction for day-0 model support, faster cold starts, and the Model Runner V2 execution engine — the Q3 2026 production upgrade path.
Deploy Ling 3.0 Flash with vLLM
Run InclusionAI's 124B MoE reasoning model locally with vLLM — 5.1B active parameters makes it feasible on a single 80GB GPU.
Deploy agentgateway on Kubernetes for MCP and A2A Traffic
Set up Solo.io's Rust-based AI gateway on Kubernetes to manage MCP tool routing, A2A agent communication, cost controls, and LLM inference proxying.
Run Muse Glimmer 30B Locally with llama.cpp and DFlash
Deploy Meta's open-weight 30B agentic model on a single consumer GPU using llama.cpp with DFlash speculative decoding for 2-3x throughput improvement.
Deploy smolagents with Docker Sandboxing
Run HuggingFace's code-first agent framework with Docker-based code execution isolation — the safest local setup for agents that write Python as actions.
Migrate MCP Servers to the 2026-07-28 Stateless Spec
The MCP specification went stateless on July 28, 2026. This recipe walks through migrating your MCP servers from the old stateful protocol to the new stateless HTTP model.
Deploy vLLM Semantic Router for Mixture-of-Models Routing
Stand up the vLLM Semantic Router in front of your LLM backends to automatically route requests by intent, cache semantically, and filter unsafe inputs.
Deploy Multi-Model Routing with OpenRouter and LiteLLM
Set up a cost-optimized routing layer that sends routine tasks to cheap models and complex reasoning to frontier models — with automatic fallback and cost tracking.
Cloudflare Temporary Worker Deployment for Agents
Let your AI agent deploy a Cloudflare Worker, D1 database, or KV store with zero signup — a 60-minute temporary account for prototyping, previews, and human handoff.
Self-Host Letta for Stateful Agents with Memory
Deploy Letta (formerly MemGPT) on your own infrastructure to run stateful AI agents with persistent memory blocks, archival storage, and self-editing context.
Deploy a RAG Evaluation Pipeline with Ragas
Set up an automated evaluation pipeline that scores your RAG system on faithfulness, answer relevance, and context precision — so you know if changes help or hurt.
Deploy appCrane as a Self-Hosted Agent PaaS
Self-hosted platform for apps your AI agents build and deploy — agent-first over MCP, with enterprise SSO, per-user audit, and a secret hard-wall.
Deploy Ferrogate as a Self-Hosted AI Gateway
Stand up a Rust-based AI gateway with provider routing, virtual API keys, budget enforcement, and MCP tool execution — a high-performance alternative to LiteLLM or Portkey.
Build an Agent Test Harness with Pytest
Treat your agent's behavior like production code — write repeatable tests that catch regressions before they reach users. No more "it worked yesterday" debugging.
Deploy vLLM as an OpenAI-Compatible Server
Stand up vLLM with the OpenAI-compatible API so agents and tools can talk to open-weight models through a familiar base URL.
Open WebUI + Ollama for a Local Chat Front Door
Give non-terminal users a clean browser UI on top of local Ollama models without sending chats to a third-party SaaS.
LiteLLM Proxy for Multi-Provider Routing
Put one OpenAI-compatible proxy in front of OpenAI, Anthropic, Gemini, and local endpoints so agents can switch models with config, not code rewrites.
VS Code + Continue with Local Ollama Models
Set up the Continue extension in VS Code to autocomplete, chat, and edit with local Ollama models so your code never leaves your machine.
Deploy n8n with Local LLM Tools
Self-host n8n workflow automation and connect it to local Ollama models for private, agent-driven automations.
Self-Host SearXNG for Agent Web Search
Run a private metasearch engine that your agents can query without sending searches to commercial providers.
Run Qwen3.6-27B on vLLM with NVFP4
Production recipe for serving Qwen3.6-27B efficiently on NVIDIA hardware using vLLM and NVFP4 quantization. Cut VRAM and boost throughput without major accuracy loss.
Run Ollama on macOS with Apple Silicon
Install Ollama on M1/M2/M3 Macs with Metal GPU acceleration. The easiest path to local LLMs for macOS developers.
Build Your First OpenClaw Agent
Deploy a local OpenClaw agent on Linux or macOS, connect it to Ollama, and run your first tool-augmented task.
Deploy Hermes Agent
Install Nous Research's provider-agnostic, self-improving agent and connect it to messaging, memory, and skills.
VS Code + Continue for Local AI Coding
Set up the most popular local-first coding extension. Connect VS Code to Ollama for autocomplete, chat, and tab completion without sending code to the cloud.
Use Cline with a Local Ollama Model
Run the Cline coding agent inside VS Code, sending every prompt to a local model so your code never leaves your machine.
Docker Compose Full AI Stack
One command to deploy Ollama, Open WebUI, and Nginx with SSL. The production-ready foundation for self-hosted AI infrastructure.
Pilot Microsoft Scout
Set up Microsoft's always-on Autopilot agent for desktop and Microsoft 365, with guardrails for safe piloting.
Rent a Cloud GPU on RunPod
Get NVIDIA GPU compute by the hour without buying hardware. Deploy Ollama on cloud GPUs for local-privacy inference with cloud convenience.
Run Ollama on Ubuntu 24.04 with NVIDIA CUDA
Install Ollama on Ubuntu with GPU acceleration — the standard foundation for local agents and LLMs.
Deploy Open WebUI with Ollama
Run a self-hosted ChatGPT-style interface that talks to your local Ollama models and any OpenAI-compatible API.
Self-Hosted pgvector for RAG Applications
Add persistent memory to your AI agents with PostgreSQL and pgvector. The essential database layer for retrieval-augmented generation.
Deploy Ollama with Docker
Run Ollama in a GPU-enabled container with persistent model storage and a stable REST API endpoint.
Why these recipes are different
Tested paths
Every recipe follows a real install sequence. We note the versions that worked and the commands that failed.
Copy-paste ready
Commands are formatted to run as-is. No hand-waving about 'figure out the paths yourself.'
Failure modes included
We list the errors you are likely to see and the exact fix, not just the happy path.