OmniscioGuides
Claude Code guides, sorted by what you are trying to do
Every published guide in the guides section, in one place, grouped by the job in front of you: running several agents at once, understanding how a feature works, and clearing an error message. Each one names its own sources and the day it read them.
By the Omniscio team. Updated .
How this page was written
Each line under a title restates what that guide covers; the evidence behind it, and the date it was read, is inside the guide. A guide that is written but not yet published is left out of this index until it goes live.
Running several agents at once
What changes when one agent becomes five, and what holds them.
How to run Claude Code agents in parallel
The three things to reach for (subagents, agent view and worktrees), with what each one does, the keys that drive them, and the ceiling you hit first.
How to run multiple Claude Code sessions at once
Every way to hold several sessions at once, from a second terminal window to a manager, with the steps and the traps.
What holds 10 to 20 coding agents at once?
What breaks first at that scale, which harness to run, and how four tools answer the question of what holds them all.
What is AI coding agent orchestration?
The five jobs every setup does whether or not anyone chose to, and how much of that work each of Anthropic's five routes holds and leaves to you.
What has to be on an AI agent dashboard?
The six things a dashboard has to show, six checks that tell it from a session list, and what the built-in agent view leaves out.
How Claude Code works
One feature at a time, from what it is to how to set it up.
The Claude Code context window
What the window holds, how big it is on each model, how to read /context and the status line, and what happens when it fills up.
Claude Code skills, and how to add one
What a skill is, what one costs you before you ever use it, how to write and install one, and how it relates to slash commands.
Claude Code subagents, and when to use one
What the built-in subagents do, how to write your own in one small file, and the point where a subagent starts costing more than it saves.
Claude Code plugins
What a plugin holds, how the marketplace and the three install scopes work, where the files land, and why one can install cleanly and still put nothing in your slash menu.
Add an MCP server to Claude Code
The exact claude mcp add commands for HTTP, SSE and local servers, what each scope stores and where, and how to check, list and remove one.
AGENTS.md in Claude Code
What Claude Code reads from AGENTS.md, and the one rule that makes it skip yours in silence.
CLAUDE.md and Claude Code memory
Where the four kinds of CLAUDE.md file live, what earns a place in one, and the four reasons a file you wrote seems to have no effect.
The Claude Code sidebar
How the Code tab's sidebar groups, sorts and archives sessions, which of the two layouts you have, and where an archived session goes.
Claude Code hooks
Every hook event the documentation lists, where the configuration lives, three hooks you can copy, and what the exit code decides.
Claude Code mods
What a mod is, how it differs from plugins, hooks and skills, and how to build one.
Claude Code plan mode
How plan mode works, what it blocks and what it lets through, and when planning first is cheaper than editing.
Claude Code checkpoints and rewind
What a checkpoint captures, every action in the rewind menu, the five things a rewind does not cover, and how it differs from git.
The Claude Code status line
How to set up the status line, the fields its JSON contains, and why it can be blank or show the wrong token count.
Claude Code prompt caching
How the prompt cache is organized, the two cache lifetimes, what breaks it, and how to watch your hit rate.
Claude Code token usage
The four places your own token numbers appear, the habits that shrink them, and why a long session costs more than it looks.
The Claude Code sandbox
How to switch sandbox mode on, why it breaks git, gh and the network, and how to take one tool out of it.
Claude Code headless mode
The print-mode flags that matter in a script, the three output formats, and why a claude -p run hangs, asks for permission or prints nothing.
Claude Code in tmux
The settings that keep Claude Code working in tmux, the scrolling and send-keys routes, and where pane work stops paying off.
Fixing the errors Claude Code prints
Each of these starts from the message itself, word for word, then the causes behind it.
Claude Code error messages explained
The messages Claude Code prints when something stops it, word for word, with what each one means and the steps that clear it.
Fix "command not found: claude"
Every cause Anthropic documents for an installed Claude Code your shell cannot find, with the PATH check for each shell and the cases that only look like it.
Claude Code login errors
Ten sign-in errors, quoted from Anthropic's documentation word for word, with what each one means and the steps that clear it.
Claude Code failed to load session
Which part of the desktop app shows that message, the three causes Anthropic gives, and what to try for each.
Claude Code error editing file
How the Edit tool matches text, and every cause behind a failed match, with the fix for each.
Which one first?
If you have just installed Claude Code, read the context window and skills: those two explain most of the surprises in a first week. If you are already running one agent and want two, start with parallel agents. If something has stopped working, the error message index is the fastest way in.
Run every agent in one window
Omniscio is the desktop app we make for the problem the first group of guides describes: several agents at once, each in its own session, all visible in one list. Any model, any harness — Claude Code, Codex, OpenCode, Devin, Pi, and Hermes all supported.