Commands
Codeep provides over 40 slash commands for common development tasks. Type /help inside Codeep to see the full list at any time.
★
Commands can be typed at the prompt or chained. For example:
/diff then ask Codeep to explain the changes.Agent & Execution
| Command | Description |
|---|---|
/agent <task> | Run autonomous agent on a task |
/agent | Enter interactive agent mode |
/plan <task> | Generate a numbered plan first — review before /go executes |
/go | Execute the pending plan from /plan |
/agents | List sub-agents the agent can delegate to (planner / researcher / reviewer / tester, or your own in .codeep/agents/) |
/undo | Undo the last agent action |
/undo-all | Undo all actions from the last session |
/history | View action history for this session |
Git
| Command | Description |
|---|---|
/diff | Review uncommitted changes with AI analysis |
/diff --staged | Review staged changes |
/commit | Generate a semantic commit message |
/git-commit <msg> | Commit with a specific message |
/review | AI review of unstaged git changes (not full codebase) |
/review --staged | AI review of staged git changes |
/review --static | Static analysis — changed files, or full src/ if clean |
/review <file> | Static analysis of specific file(s) |
Session Management
| Command | Description |
|---|---|
/sessions | Browse and resume past sessions |
/new | Start a new session |
/rename <name> | Rename the current session |
/search <term> | Search within the current session |
/recall <query> | Search across ALL saved sessions, ranked by relevance + recency |
/recall <query> --resume | Load the top-matching session directly (skip the picker) |
/recall <query> --summarize | LLM recap of what you accomplished across matching sessions |
/compact [keepN] | AI-summarize older messages to free up context (keeps last N, default 4) |
/checkpoint [name] | Snapshot the session (conversation + provider/model + git HEAD) |
/checkpoints | List saved checkpoints for this workspace |
/rewind <id> | Restore a checkpoint (files via git, see hint after rewind) |
/checkpoint delete <id> | Delete a saved checkpoint |
/export [md|json|txt] | Export current session |
Context & Files
| Command | Description |
|---|---|
/add <path> | Add a file or directory to context |
/drop [path] | Remove a file (or all) from context |
/apply | Apply file changes from the last AI response |
/paste | Paste clipboard content into the conversation |
/context-save | Save current conversation context |
/context-load | Load a saved conversation context |
/context-clear | Clear the saved context |
Settings & Profiles
| Command | Description |
|---|---|
/settings | Open interactive settings menu |
/provider | Switch AI provider |
/model | Switch model for current provider |
/model pull <name> | Pull an Ollama model (local Ollama only) |
/thinking [auto|low|medium|high|max] | Set the thinking / reasoning-effort tier — how hard the model reasons. Auto = each model's default; tiers are clamped per provider+model so an unsupported value is never sent. Works on Anthropic Opus/Sonnet, OpenAI GPT-5.x, Google Gemini 3, DeepSeek V4, and Z.AI GLM-5.2. |
/effort | Alias for /thinking |
/lang | Set response language |
/login | Add or update an API key |
/logout | Remove an API key |
/keysync [on|off|status] | Toggle syncing API keys to codeep.dev — off by default; keys stay in your OS keychain unless you opt in |
/profile save <name> | Save current provider, model and settings as a profile |
/profile load <name> | Load a saved profile |
/profile list | List all saved profiles |
/profile delete <name> | Delete a profile |
/openrouter | Tune OpenRouter routing: sort by price/throughput/latency, allow training, disable specific upstreams |
/personality | List or switch agent tone (concise / verbose / security / senior-reviewer / junior-mentor / ship-it) |
/personality <name> | Activate a personality (built-in or user-authored). /personality off to clear. |
/me | Your user profile — adapts the agent to you (reply language, style, stack). Injected on every run, every surface. |
/me init [project] | Scaffold a profile template (global ~/.codeep/profile.md, or project .codeep/profile.md) |
/me learn [on|off] | Learn durable prefs from this session now; on/off toggles opt-in auto-learn. /me forget clears it |
/me sync | Push your profile to the codeep.dev dashboard (and pull on a fresh machine) |
Project Intelligence
Initialize a project, scan it once to give the AI deep context about its structure, frameworks, and conventions. Use /memory to add your own notes — they are included in every AI and agent conversation and synced to the dashboard.
| Command | Description |
|---|---|
/init | Initialize project — creates .codeep/ folder |
/scan | Scan project and cache intelligence for AI |
/scan status | Show last scan info (age, file count, project type) |
/scan clear | Clear cached intelligence |
/memory <note> | Add a custom note to project intelligence |
/memory list | Show all notes with indices |
/memory remove <n> | Remove note by index |
/memory clear | Clear all notes |
★
Example notes:
/memory Always use pnpm, never npm or /memory Main entry point is src/renderer/main.tsUtility
| Command | Description |
|---|---|
/help | Show all available commands |
/status | Show current status |
/cost | Show session token usage and estimated API cost per model (with prompt-cache savings) |
/stats | Detailed session view — per-model breakdown, prompt-cache summary, and the per-1M pricing table |
/commands | List user-authored slash commands from .codeep/commands/*.md |
/hooks | List installed lifecycle hooks from .codeep/hooks/<event>.sh |
/mcp | List MCP servers connected to this session and the tools they expose |
/tasks | List pending tasks (numbered; tagged by project name when listed outside a project) |
/tasks add <title> [--bug|--feature] [--desc <text>] | Create a task on the dashboard — optional type (bug/feature, default task) and description |
/tasks done <n> | Mark task #n as done |
/tasks delete <n> | Delete task #n permanently |
/sync | Sync learning preferences and profiles to codeep.dev (multi-device) |
/sync learning | Sync only learning preferences |
/sync profiles | Sync only saved profiles |
/clear | Clear the chat |
/multiline | Toggle multi-line input mode |
/copy [n] | Copy code block to clipboard |
/version | Show Codeep version |
/update | Check for updates |
/insights | Activity summary over the last 7 days — runs, files, tools, projects (use --days N to widen) |
/insights --days 30 | Same, with custom window |
Skills & Extensions
Skill bundles, MCP servers, and lifecycle hooks extend the agent. Full guides: Skills, MCP.
| Command | Description |
|---|---|
/skills | List the 50+ built-in skills (and search them) |
/skills bundles | List structured skill bundles in .codeep/skills/ |
/skills create-bundle <name> | Scaffold a new SKILL.md bundle |
/skills show <name> | View a bundle's full SKILL.md |
/skills browse [query] | Browse the public marketplace at codeep.dev/skills |
/skills install <owner>/<slug> | Install a published skill into the project |
/skills publish <name> [--public] | Publish a bundle to the marketplace |
/skills unpublish <owner>/<slug> | Remove a published skill |
/mcp | List connected MCP servers and their tools |
/mcp browse [id] | Browse the 12-server marketplace |
/mcp install <id> [args] | Install a marketplace server |
/mcp add <name> <command> | Add a custom MCP server |
/mcp resources / read <uri> | Explore + read MCP resources |
/mcp prompts / prompt <s> <n> | Explore + materialize prompt templates |
Session Checkpoints
Before a risky multi-step refactor (or when the agent starts going off the rails), save a named snapshot of the conversation. Restore it later if things go sideways. File rollback is delegated to git — Codeep prints the exact git checkout command after rewind.
/checkpoint before-auth-refactor # save snapshot
/checkpoints # list saved checkpoints
/rewind ck-2026-05-18-a1b2c3d4 # restore conversation + provider/model
/checkpoint delete ck-... # remove a snapshotSnapshots include: full conversation history, current provider + model, the list of files the agent touched in this session, and (if cwd is a git repo) the short SHA of HEAD at checkpoint time. Stored under .codeep/checkpoints/<id>.json in your project.
★
Checkpoints don't snapshot file content — by design. Use
git stash or the suggested git checkout <sha> -- <file> printed after rewind to roll files back.Lifecycle Hooks
Drop executable shell scripts in .codeep/hooks/<event>.sh and Codeep runs them at the relevant moment during a session. Use them to enforce policy (block writes to certain paths), auto-format edits, run lint gates before commit, or log every tool call to your own pipeline.
| Event | When | Blocking? |
|---|---|---|
pre_tool_call | Before every agent tool call (built-in or MCP) | Yes — non-zero exit blocks the call |
post_edit | After write_file / edit_file succeeds (built-in only) | No (advisory) |
on_error | When any tool call fails (built-in or MCP) | No (logging) |
pre_commit | Before /commit stages anything | Yes — non-zero exit blocks the commit |
Each hook receives these environment variables:
CODEEP_HOOK_EVENT # event name
CODEEP_WORKSPACE # workspace root absolute path
CODEEP_SESSION_ID # current session id
CODEEP_HOOK_TOOL # tool name (pre_tool_call, on_error, post_edit)
CODEEP_HOOK_PARAMS # JSON-encoded tool parameters (pre_tool_call, on_error)
CODEEP_HOOK_FILE # absolute file path (post_edit only)Example — auto-format on every edit:
#!/bin/bash
# .codeep/hooks/post_edit.sh
prettier --write "$CODEEP_HOOK_FILE" 2>/dev/null⚠
Hooks are arbitrary shell that runs whenever an agent tool fires. Codeep shows a warning banner on session start if hooks are installed — for the same reason we don't read hooks from a global
~/.codeep/hooks/ directory (only project-scoped, so cloning a hostile repo at least flashes the warning before scripts can run).Custom Slash Commands
Drop a Markdown file in .codeep/commands/<name>.md (project-scoped) or ~/.codeep/commands/<name>.md (global, all projects) and Codeep exposes it as /<name>— in the TUI, the VS Code extension, and any other ACP client (Zed). The file's body becomes the user prompt; the agent handles it normally.
Example: a security review template
Save the following as .codeep/commands/sec-review.md:
---
description: Security review of a file
aliases: [sec, secrev]
---
Please perform a thorough security review of {{args}}. Look for:
- AuthN/AuthZ bypasses, SQL/NoSQL injection, XSS, CSRF
- Path traversal, insecure deserialization
- Hardcoded secrets, race conditions
Report findings ordered by severity.Then call it as /sec-review src/api/login.ts (or /sec via the alias).
Placeholders
| Placeholder | Resolves to |
|---|---|
{{args}} | Full argument string |
$ARGUMENTS | Full argument string (Claude Code compatibility) |
{{arg1}}, {{arg2}}, … | Positional arguments (1-indexed) |
★
Project files shadow global files with the same name, so a per-repo
.codeep/commands/deploy.md can override a global one. Run /commands to list everything available in the current workspace.Image Paste
Press Ctrl+V to paste an image directly from the clipboard into the conversation. Codeep will automatically detect the image and send it to a vision model for analysis.
★
Requires a Z.AI or MiniMax API key. On macOS, copy a screenshot with
Cmd+Shift+4 then press Ctrl+V in Codeep.