All built-in aegiscode commands in one place. Click ⧉ Copy on any example, paste it straight into your terminal session.
33 commands · organized by what they do
Run these before you ever type a slash command — install, log in, and (optionally) turn on cross-session memory.
Requires Node.js 22+. Installs the aegis command globally.
npm install -g aegiscodeNative Electron app — same engine, same memory, with a split-pane terminal + shell. Linux, macOS, Windows.
wget https://dl.aegiscloud.org/aegiscode-gui.deball platforms: aegiscloud.org/aegiscode#download
First run with no API keys configured launches this automatically — pick a provider, paste a key, done. Run it manually any time to redo setup.
aegis --initLog in to aegiscloud.org — opens a browser for Google sign-in. This is what unlocks /cloud sync and the €2/mo memory subscription.
aegis loginaegis login --passwordaegis logoutAlready paying for Claude Code Pro/Max? Use that subscription instead of a pay-per-token Anthropic API key. Requires the claude CLI installed and on PATH.
claude setup-tokenaegis login --claude-proTurn on persistent cross-session memory (€2/mo) — subscribe at aegiscloud.org, check your inbox for the token, then activate inside a running session.
/memory activate <token>subscribe first at aegiscloud.org/subscribe — full details under Memory & Cloud below
Manage the current conversation — clear it, compress it, check what it's costing you.
Show all available commands, or detailed help for one specific command.
/help/help modelWipe the current chat history and start fresh, same session ID.
/clearManually compress the conversation into a summary to free up context window — happens automatically near the token limit too.
/compactShow session info — current model, permission mode, message count, context usage.
/statusShow a token usage graph for the session and an estimated dollar spend.
/tokensShow the installed aegiscode version.
/versionSwitch models, let aegiscode pick one automatically, and control how much it asks before acting.
Show the interactive model switcher, jump straight to a model by ID, list configured models, or add/remove a custom one.
/model/model claude-sonnet-4/model add openrouter-mixtral "Mixtral 8x7B" mistralai/mixtral-8x7b-instruct https://openrouter.ai/api/v1 sk-or-...usage: /model [id] | /model add <id> <name> <model> <baseURL> <apiKey> | /model remove <id> | /model list
Auto-pick the cheapest model that can handle each message based on task complexity, and learn from aborted responses over time.
/router on/router set simple deepseek-chat/router statsusage: /router [on|off|set <simple|medium|complex> <modelId>|stats]
Set Claude's extended-thinking effort level for harder problems.
/effort highusage: /effort [off|low|medium|high|max]
Toggle the tool-call confirmation prompt, globally or for one specific model.
/confirm offusage: /confirm [on|off] [model-id|claude|deepseek|all]
Toggle YOLO mode — every tool call (file edits, shell commands) gets auto-approved, no prompts.
/yolo onusage: /yolo [on|off]
Put more than one model to work on the same problem — in parallel, in a vote, or in a structured argument.
Orchestrate multiple AI agents on a complex task, each handling a piece.
/multi build a REST API with auth and testsusage: /multi <task> [--save-as <name>] [--template <id>]
Same as /multi, with every agent's tool calls auto-approved.
/multiyolo refactor the entire auth moduleusage: /multiyolo <task>
Decompose an app idea into components and build everything in parallel — DeepSeek for backend, GPT-4o for frontend, Llama for tests, all writing real files at once.
/build a Flask web app with login and dashboardusage: /build <description>
Send a question to every configured model simultaneously and surface the majority answer, dissents included.
/council is a state machine overkill for this form?usage: /council <question>
Run a structured, multi-round debate — each model sees the others' arguments and pushes back or refines.
/debate monorepo vs polyrepo for this team --rounds 3usage: /debate <topic> [--models id1,id2] [--rounds N] [--format debate|discussion|qa|panel]
Research a topic via multi-agent deliberation instead of a single model's first answer.
/research what's the best way to handle websocket reconnection?usage: /research <question>
Clone any website's structure and styling into a local project using DeepSeek.
/clone https://example.com --name my-cloneusage: /clone <url> [--name <project-name>]
Cross-session semantic memory and cloud sync — the part that needs a subscription.
Activate, inspect, or manage persistent semantic memory — the AI remembering your stack and past decisions across sessions.
/memory activate <token>/memory stats/memory clearusage: /memory [activate <token> | load <url|path> | upload | clear | stats]
Manage AEGIS Cloud sync — uploads every session to aegiscloud.org/dashboard, browsable from any browser.
/cloud activate/cloud statususage: /cloud [status | key <api_key> | activate | deactivate]
Show current subscription and billing status.
/billingTheme, skills, hooks, MCP servers — the parts that make aegiscode fit your setup specifically.
Show or switch the UI theme.
/themeusage: /theme [theme-name]
Toggle whether extended-thinking blocks render expanded or collapsed.
/thinkingCopy the last code block (or a specific one) straight to your clipboard.
/copy/copy 2usage: /copy [n | last | list | raw]
List discovered skills (project + user-level SKILL.md files) or refresh the index.
/skillsusage: /skills [name|refresh]
View configured hooks and their status.
/hooksusage: /hooks [status|list]
Show connected MCP server status and the tools each one exposes.
/mcpusage: /mcp [tools|<server-name>]
No commands match that filter.