Slash Commands

Instant access to powerful agent behaviors — type / and choose exactly how you want the agent to work

What it is

Slash Commands are one-keystroke shortcuts that change how the agent approaches your task. Type / in the chat input to open the command menu, select a command, and the agent operates with that mode active — planning before acting, thinking harder, fact-checking claims, storing to memory, and more.

They are the fastest way to move from free-form chat to directed, structured agent behavior.

How to Use

1. Type / at the start of your message in the chat input
2. The command menu appears above the input
3. Continue typing to filter: /pl shows /plan, /plot, /playback
4. Arrow keys navigate, Enter or Tab selects, Escape dismisses
5. Click any command to select it directly

After selecting, the command's instruction prefix is injected into your message. Describe your task as usual — the command modifies how the agent handles it.

Built-in Commands

Planning & Thinking

Implementation

Quality & Review

Output & Memory

Examples

Using /plan

/plan Build a Stripe webhook integration that handles subscription lifecycle events

The agent responds with a detailed implementation plan covering every step — database schema, webhook handlers, event types, error handling, testing approach — and waits for your approval before writing any code.

Using /think-hard

/think-hard Should we build our own auth or use Clerk?

The agent works through both options systematically — cost, maintenance, security, time-to-market, lock-in risk, team capability — before presenting a reasoned recommendation.

Using /review

/review [paste code or describe what to review]

The agent reviews for bugs, edge cases, security vulnerabilities, performance problems, and code quality. Each finding is specific with a concrete suggested fix.

Using /remember

/remember Our fiscal year runs February to January. We use Linear for sprint tracking. Our primary database is Convex on the production deployment artful-fly-108.

The agent stores this to your memory block. All future interactions — including scheduled tasks and workflows — have access to this context automatically.

Using /research

/research What are the main regulatory requirements for deploying AI agents in financial services in the US and EU?

The agent searches multiple sources, cross-references findings, identifies consensus and gaps, and produces a structured, cited report.

Custom Commands

Create your own slash commands to capture your team's most-used workflows.

Creating a Custom Command

1. Go to Dashboard → Agents → Slash Commands
2. Click New command
3. Give it a name (e.g., competitor-brief)
4. Write a description
5. Write the prompt template — use $ARGUMENTS where your task text should be inserted

Example: /competitor-brief

Command: competitor-brief
Description: Generate a structured competitor brief

Prompt Template:
Research the following company thoroughly for competitive intelligence purposes.
Include: company overview, product features, pricing, ICP, key differentiators,
known weaknesses, recent news, and a recommended approach for competing against them.
Output as a structured brief in the Document panel.

Company: $ARGUMENTS

Now typing /competitor-brief Salesforce triggers that full research prompt automatically.

Managing Commands

Visit Dashboard → Agents → Slash Commands to:

• View all built-in commands with their prompt templates (expandable)
• Create new custom commands
• Edit your custom commands
• Delete custom commands you no longer need

Built-in commands are read-only but can be inspected — click any command row to see the full prompt template it injects.

Tips

Use /plan for any task you're not fully sure about. Getting a plan back before implementation lets you catch misunderstandings early — before the agent spends 20 minutes building the wrong thing.

Use /remember for anything you say more than once. If you've explained your tech stack, naming conventions, or team preferences before, tell the agent to remember them. You shouldn't have to repeat yourself.

Chain commands in sequence. For complex work: start with /plan, approve the plan, then follow up with /implement. Or: /research the landscape, then /analyze what you've found, then /build the solution.

Build a custom command for every recurring workflow. If your team does the same type of task repeatedly — competitor briefs, PR descriptions, sprint recaps — create a command that captures the template and shares it across your workspace.

Common Questions

Do slash commands consume extra credits? No. Commands modify how the agent processes your request but don't add overhead billing.

Can I use slash commands in scheduled tasks? Not directly — scheduled tasks use full natural language instructions. But you can copy a command's prompt template directly into a scheduled task description.

Can my whole team use the same custom commands? Custom commands are currently stored per user in the browser. Workspace-level shared commands are on the roadmap.

Can I override a built-in command? Not directly — built-in commands cannot be deleted. But you can create a custom command with a similar name (e.g., plan-v2) that uses your preferred template.

What's the difference between /plan and just asking "make a plan first"? Functionally similar. /plan is a one-keystroke shortcut that ensures the exact planning instruction is applied consistently every time, without you having to remember the phrasing.