Master AI tools, agents, and workflows. Become AI-enabled.

We show how real AI systems are set up and run, one at a time. Then we sell the guides behind them, so you can do the same.

Anyone who wants to run their own AI agents for real, technical or not.

A news reposter. A generic AI content farm. A prompt-pack store. An agency pushing bespoke services. An AI avatar presented as a real person.

Claude Code is an agentic coding tool that reads your codebase, edits files, runs commands, and integrates with your development tools. It is available in your terminal, your IDE, a desktop app and the browser.

An MCP server is a program that provides context to MCP clients. MCP, the Model Context Protocol, is an open-source standard for connecting AI applications to external systems. A server can run on your own machine or remotely.

Ollama lets you run open models on your own computer, or use them in the cloud. Run locally, it serves an HTTP API on 127.0.0.1, port 11434 by default.

A self-hosted AI agent is one whose agent loop you run on your own infrastructure, instead of a vendor running it for you. Anthropic's documentation draws this line: you can self-host the Agent SDK on your own servers, or use Managed Agents, where Anthropic hosts the agent loop.

MCP is a standard protocol between an AI application and the servers it connects to. It uses JSON-RPC 2.0, so the application can discover a server's tools and call them. A tool is an executable function, and the documentation lists API calls among its examples.

A runbook is a detailed how-to guide for completing a commonly repeated task or procedure.

A playbook deals with the overarching responses to larger issues and events. A runbook outlines the smaller individual tasks. The playbook accounts for the big picture.

A common structure has seven sections: service overview, build information, instructions for deploying the software, instructions for common tasks, a pager playbook, disaster recovery plans and a service level agreement. Keep it clear, specific, structured the same way every time, and test it before you rely on it.

Yes, Ollama documents it. Run `ollama launch claude`, or point Claude Code at `http://localhost:11434`. For local models, set a 64k+ context window and pick a model with tool support.

Install Claude Code, then set `ANTHROPIC_BASE_URL=http://localhost:11434` and `ANTHROPIC_AUTH_TOKEN=ollama`. Run `claude --model` with a local model.

Run locally, Ollama says it doesn't see your prompts or data, and it listens on 127.0.0.1 by default. For its cloud models, Ollama says it does not store or log your content and never trains on it.

Running models on your own hardware is always unlimited, and the Free plan costs $0. Cloud models use pay-as-you-go credits, and paid plans exist above the Free plan.

CLAUDE.md is a markdown file you add to your project root that Claude Code reads at the start of every session. Use it for coding standards, architecture decisions, preferred libraries and review checklists. Claude treats it as context, not as enforced configuration.

Monitoring is collecting, processing, aggregating and displaying real-time quantitative data about a system. Google's SRE guide tracks four signals: latency, traffic, errors and saturation. With the Claude Agent SDK, set OpenTelemetry variables so each query exports traces, metrics and logs to your collector; prompt text and tool inputs are not included by default.

With the Claude Agent SDK, each agent runs as its own process, so you host it in a container with outbound HTTPS access to the API. Pick a pattern by how long the container lives: ephemeral, long-running, hybrid or multi-agent. A reasonable starting point is 1 GiB of RAM, 5 GiB of disk and 1 CPU per agent.

OWASP's 2025 Top 10 for LLM applications names the main risks, including prompt injection, sensitive information disclosure and excessive agency. For an agent, Anthropic's documentation advises running it in a sandboxed container, keeping tool credentials out of the agent environment, and sending outbound calls through a proxy that adds the credentials.

Each product page shows its price, and the Polar checkout shows it in the currency displayed. Polar sells the products as merchant of record, so it works out and collects the sales tax or VAT that applies to your purchase. The free guides cost nothing.

Once Polar confirms your payment, the product is made available through Polar as a signed, personal download link. Keep a copy of your files. If you cannot access something you bought, write to hello@stellarwind.ai.

No. Stellarwind products are digital downloads, available as soon as your payment is confirmed, so they are not refundable. At checkout you ask for immediate access and confirm that you give up your right of withdrawal, as the terms of sale set out. Polar, which takes the payment, handles any refund the law still requires.

Each purchase gives you a personal, non-exclusive and non-transferable licence to use the product for yourself or within your own business. You may not resell, share or redistribute it, or include it in a product or service you offer to others. Ownership stays with Stellarwind.