coachix.
Private Assistant
Coachix lets you shape a personal AI assistant around your knowledge, privacy boundaries, trusted ideas, and evolving ways of working. It is not just a chat box; it is a workspace where the assistant can become more yours over time.
Open Personal AI AssistantChoose what role it plays, which tools it may use, what skills it can run, and what tone or workflow feels natural to you. The assistant becomes specific to your tasks instead of starting every day from a generic chatbot.
Design the assistant Private knowledgeSave notes, documents, decisions, and source material as private context. The assistant can work from the information you provide while personal, business, and research material stay separated from disposable public chat.
Protect knowledgeGive the assistant principles, philosophies, decision rules, and review habits you actually agree with. When it plans, writes, checks, or decides whether to ask before acting, it can follow the thinking patterns you trust.
Guide its judgment Self-evolving companionFeedback, better skills, clearer knowledge, and refined rules can make the next task better than the last. Over time, your assistant remembers what works, adapts to your patterns, and grows alongside you.
Grow togetherComplete Guide
In this guide, you'll learn how to build a private AI agent from scratch without treating the agent as a single prompt. Start with Philosophy: the operating principles that tell the assistant what good judgment looks like, when to ask, and when to stop. Then build the Knowledge layer, where notes, documents, decisions, and source material become context the agent can retrieve instead of guessing from memory. After that, add MCP tools so the assistant can reach the systems it needs through explicit, inspectable permissions rather than vague automation. Skills come next: reusable procedures for research, writing, deployment, recovery, review, and handoff, so the agent can repeat work in a dependable way. Finally, choose LLM routes that match the job: a stronger planner for hard reasoning, a faster worker for routine steps, and a review path for decisions that affect money, users, security, or public content. The goal is not to build a louder chatbot. The goal is to assemble a private assistant that knows its role, uses bounded tools, works from your material, improves through feedback, and leaves a trace you can review before trusting the result. Use the modules below as the build order, moving from judgment to knowledge, tools, skills, and model choice.
Use the Building AI Agent workflow around one job first: decide what the assistant should plan, what it may execute, and what result needs human review. A focused first job keeps philosophy, knowledge, MCP tools, skills, and model choice aligned.
Use the Building AI Agent workflow to shape memory from trusted notes, source documents, decisions, and examples. Treat knowledge as a private working set the assistant can retrieve, cite, and update after feedback instead of guessing.
Use the Building AI Agent workflow to add MCP tool access only when a task needs it. Give each tool a clear permission boundary, expected input, and review point so browser work, files, deployments, and recovery stay inspectable.
Use the Building AI Agent workflow to package skills as repeatable recipes for research, writing, coding, deployment, and repair. Skills turn one good workflow into a reusable procedure that the assistant can run again without losing context.
Use the Building AI Agent workflow to create review loops before trust. Choose a stronger model for planning, a faster route for routine execution, and a second pass for decisions involving money, users, security, or public content.
Use the dedicated guide for the long-tail topic: purpose, knowledge, MCP tools, skills, review loops, and model routes for building AI agent workflows.
Model and release pages live under the LLM route so the Coachix homepage stays focused on the official brand and private assistant workspace.
It organizes private AI agent setup notes, reasoning references, and machine-readable entrypoints for builders.
It is for builders and small teams wiring agents to tools, memory, browser actions, and self-hosted workflows.
Start with resources, then move through recovery, reasoning, self-correction, and Existence Theory when a workflow needs a stronger operating frame.
coachix is primarily a public reference site for private AI agent setup and reasoning. The pricing page clarifies when public pages are enough and when a team might ask for implementation support around agent memory, recovery behavior, workflow boundaries, or machine-readable catalog integration.
Use the public resources, sitemap, and llms.txt first. Paid support should only be considered when the team has a concrete agent reasoning problem, public-safe workflow evidence, and a clear owner for follow-up.