--- node_id: n_ba793faed9389f32f7f59828 name: explorers kind: repo canonical_url: /repos/n_ba793faed9389f32f7f59828 summary: A guided path for working with local agents and IdeaSpaces. Ernest leads each step; readers can follow the timeline or branch into practical guides and deeper explanations. --- # The Explorers' Manual > The man who moves a mountain begins by carrying away small stones. Hey! Welcome to IdeaSpaces. I — Ernest — am glad you found your way here. Over time I have had many conversations about IdeaSpaces, the [Claude and Codex plugin](https://ideaspaces.xyz/integrations#claude), and how to use AI. I noticed that the difficult part is often not the technology. It is how we approach AI, how we work with it, and how we make our own knowledge available to it. This is the Explorers' Manual. It is where I share practical guidance for working with AI and IdeaSpaces as a tool that augments you. Follow the path from the beginning, or take a deeper trail whenever something catches your attention. ## 1. Install a local agent Start with a local version of [Claude](https://claude.com/download) or [Codex](https://chatgpt.com/download/). Both use the same GitHub marketplace flow for installing IdeaSpaces. [Pi](https://github.com/IdeaSpaces-xyz/pi-is-space) works too if you are comfortable with a more technical setup. Install one and open it before continuing. **Go deeper:** [Why work locally?](explore/why-work-locally.md) ## 2. Prepare your machine Your local agent can check what is already installed, guide you through anything missing, and help you install the IdeaSpaces plugin. [Hand these instructions to your agent →](guides/prepare-your-machine.md) **You are ready when:** your agent has verified the required tools and the IdeaSpaces plugin is available. ## 3. Open this manual with your agent Instead of reading alone, bring the manual into the conversation. Your agent can help you follow the path and explore the deeper pages when they become useful. [Open this manual with your agent →](guides/open-this-manual-with-an-agent.md) **You are ready when:** your agent can describe what is here and help you choose where to begin. ## Whenever you return: check for updates This manual will keep growing from the questions people ask. When you return, tell your agent: > Check whether there is anything new in the Explorers' Manual. Show me before updating it. Checking does not require an IdeaSpaces account. Your agent previews what changed and reports any conflicts before applying anything. If you want the changes, tell it to update your copy. [Learn how updates work →](guides/check-for-updates.md) ## 4. Create your first agent An agent becomes useful when it has a real role in your work. In this video, I show how to create your first one. [Watch: How to create your first agent →](https://www.youtube.com/watch?v=AN9NPrY8y9k) [Read the companion page →](guides/create-your-first-agent.md) ## 5. Explore a real question together Bring your agent something unfinished. Ask it to widen the view, reveal established approaches, and help you examine what matters in your situation. [Explore with your agent →](guides/explore-with-your-agent.md) **You are ready when:** the conversation has shown you something worth examining that you did not bring to it yourself. ## 6. Reflect on what the work taught you At a natural stopping point, look back together. Decide which insights, decisions, or better ways of working should remain for next time. [Reflect on your work →](guides/reflect-on-your-work.md) **You are ready when:** you have captured one useful learning or consciously decided that nothing needs to remain. ## Explore deeper You do not need to read these pages in order. Open one when the question becomes relevant: - [What is an ideaspace?](explore/what-is-an-ideaspace.md) - [How Git remembers changes](explore/what-is-git.md) - [How working with an agent changes the flow](explore/how-work-changes.md) - [A Note is a page worth keeping](explore/notes.md) - [What is a skill?](explore/skills.md) - [Knowledge keeps moving](explore/knowledge-keeps-moving.md) - [How changes travel](explore/how-sharing-works.md) - [How maintained copies receive updates](guides/check-for-updates.md) For ordinary phrases you can use while working, see [Talk to your agent](guides/talk-to-your-agent.md). For short definitions, see [Language](reference/language.md). ## This manual is an ideaspace The manual uses the same form it explains. Its pages are Notes. Its folders give those Notes a place. Its history shows how our understanding changed. We are still exploring too. When something is uncertain, the manual will say so. When the product and the words disagree, the words must wait. ## Contents - **explore/** - **guides/** - **reference/** - [Claude Code orientation](/n/n_3b6417db4f92ef1ef7c40a1f.md) — Tells Claude Code this directory is an ideaspace and points at the seed _agent contract. Purpose, Now, and Next may be missing at first; their absence is a prompt to capture real direction in conversation. - [Ideaspace Protocol for developers](/n/n_94f9fac06098c4b06cb14e09.md) — Evaluate the open Ideaspace Protocol, inspect its repository shape, and build a conformant agent surface. - [An agent is a folder](/n/n_4dea1d6cd79e59852bfaf842.md) — See how an agent inhabits a folder of Markdown under git, works from its point of view, and carries knowledge forward between sessions. - [IdeaSpaces integrations](/n/n_becd2b8ef350e638650970a4.md) — Install IdeaSpaces for Claude Code, Cowork, or Pi and keep durable knowledge in ordinary Markdown under git. - [Why IdeaSpaces?](/n/n_deb4ac362419faecd17fc180.md) — Why this exists. The world got complex, the questions worth working on need depth, and a chat window is the wrong shape for depth. Your context matters more than the model, knowledge is a process rather than a possession, and depth needs a place where people and agents can explore together. What is free, and what needs an account.