Commands Reference

Complete reference for all slash commands available in B4M CLI.

Slash commands are special commands that start with / and control the CLI's behavior rather than sending messages to the AI agent.

Quick Reference

Command	Description
`/login`	Authenticate with Bike4Mind
`/logout`	Sign out and clear tokens
`/whoami`	Show current user
`/sessions`	List saved sessions
`/save <name>`	Save current session
`/clear`	Start a new session
`/rewind`	Rewind conversation to previous point
`/set-api <url>`	Connect to self-hosted instance
`/reset-api`	Reset to default API
`/api-info`	Show API configuration
`/trust <tool>`	Trust a tool
`/untrust <tool>`	Remove tool from trusted list
`/trusted`	List trusted tools
`/mcp`	Show MCP server status
`/usage`	Show credit usage and balance
`/config`	Open configuration editor
`/commands`	List custom commands
`/commands:new <name>`	Create custom command
`/commands:reload`	Reload custom commands from disk
`/project-config`	Show merged project configuration
`/sandbox`	Show sandbox status
`/sandbox:enable`	Enable OS-level sandbox
`/sandbox:disable`	Disable sandbox
`/sandbox:mode <mode>`	Set sandbox mode
`/sandbox:trust-domain <domain>`	Trust a network domain
`/sandbox:domains`	Show allowed domains
`/sandbox:violations`	Show recent violations
`/sandbox:violations:clear`	Clear violation log
`/help`	Show help
`/exit`	Exit CLI

Authentication Commands

`/login`

Initiate OAuth authentication flow with Bike4Mind.

Usage:

/login

What happens:

Displays device authorization screen
Automatically opens browser with activation URL
Code is prepopulated in the browser
You verify the code and click "Approve"
CLI detects authorization and stores tokens

Example:

> /login

╭────────────────────────────────────────────────────────╮
│                                                        │
│ 🔐 Device Authorization                                │
│                                                        │
│ Opening browser automatically...                       │
│ If it doesn't open, please visit:                     │
│                                                        │
│   https://app.bike4mind.com/activate                  │
│                                                        │
│ And enter this code when prompted:                    │
│                                                        │
│   L2JQ-CRZ9                                           │
│                                                        │
│                                                        │
│ ⠏ Waiting for user authorization...                   │
│                                                        │
│ Expires in 10 minutes                                 │
│                                                        │
╰────────────────────────────────────────────────────────╯

✓ Authentication successful!
Logged in as: john@example.com

When to use:

First time running the CLI (required for AI features)
After logging out
When tokens expire (though auto-refresh usually handles this)
Switching accounts

`/logout`

Sign out and clear authentication tokens.

Usage:

/logout

What happens:

Removes access and refresh tokens from config
Clears user information
Requires re-authentication to continue

Example:

> /logout

✓ Logged out successfully
You'll need to authenticate again to continue.

When to use:

Switching to a different account
Security: removing credentials from shared machine
Troubleshooting authentication issues

`/whoami`

Display information about the currently authenticated user.

Usage:

/whoami

Example:

> /whoami

Authenticated as:
  Email: john@example.com
  User ID: usr_abc123
  API Endpoint: https://app.bike4mind.com/api
  Token Status: Valid (expires in 89 days)

When to use:

Verifying you're logged in
Checking which account is active
Confirming API endpoint configuration
Debugging authentication issues

Session Management Commands

`/sessions`

List all saved sessions with metadata.

Usage:

/sessions

Example:

> /sessions

Saved Sessions:
  1. project-work (3 hours ago, 42 messages)
  2. research-notes (1 day ago, 18 messages)
  3. code-review (3 days ago, 27 messages)
  4. debugging-session (1 week ago, 56 messages)

Session Information Shown:

Session name
Last modified time
Number of messages
Token usage (if available)

When to use:

Finding a previous conversation
Managing saved work
Checking session history

`/save <name>`

Save the current conversation session.

Usage:

/save <session-name>

Arguments:

session-name - Name for the saved session (alphanumeric, hyphens, underscores)

Example:

> /save my-project-work

✓ Session saved as: my-project-work
Location: ~/.bike4mind/sessions/my-project-work.json

What gets saved:

Full conversation history
Tool execution results
Agent reasoning steps
Token usage statistics
Metadata (timestamp, model, etc.)

When to use:

End of a work session
Before switching tasks
Preserving important conversations
Creating checkpoints

Tips:

Use descriptive names: typescript-migration, bug-investigation
Save early and often
Sessions are deduplicated by name (overwrites existing)

`/clear`

Start a new conversation session (clears current context).

Usage:

/clear

Example:

> /clear

✓ Session cleared
Starting fresh conversation.

What happens:

Clears current conversation history
Resets context (agent forgets previous messages)
Starts a new session with empty state

When to use:

Switching to a completely different topic
Starting fresh after a long conversation
Clearing context to avoid confusion
Before working on unrelated tasks

Note: Consider using /save before /clear to preserve your work.

`/rewind`

Rewind the conversation to a previous point.

Usage:

/rewind

Example:

> /rewind

Select a point to rewind to:
  1. Before last agent response (2 messages ago)
  2. Before last 5 messages
  3. Custom...

> 1

✓ Conversation rewound to 2 messages ago
You can now continue from this point.

What happens:

Removes recent messages from conversation history
Returns to an earlier state in the conversation
Allows you to branch in a different direction

When to use:

Agent made a mistake and you want to retry
Conversation went in the wrong direction
Testing different approaches to a problem
Undoing recent exchanges

API Configuration Commands

`/set-api <url>`

Connect to a self-hosted Bike4Mind instance.

Usage:

/set-api <api-url>

Arguments:

api-url - Base URL of your Bike4Mind instance (must include https://)

Example:

> /set-api https://b4m.company.com

✓ API endpoint updated
New endpoint: https://b4m.company.com/api

You'll need to re-authenticate with this instance.

What happens:

Updates apiUrl in config
Clears existing authentication tokens
Prompts for re-authentication

When to use:

Your organization runs a self-hosted instance
Using a dedicated enterprise deployment
Testing against a staging environment

Security Note: Only connect to trusted instances you control.

`/reset-api`

Reset API endpoint to the default Bike4Mind service.

Usage:

/reset-api

Example:

> /reset-api

✓ API endpoint reset to default
Endpoint: https://app.bike4mind.com/api

You'll need to re-authenticate.

What happens:

Resets apiUrl to https://app.bike4mind.com
Clears authentication tokens
Prompts for re-authentication

When to use:

Returning to the main service after using self-hosted
Fixing misconfigured API endpoint
Troubleshooting connection issues

`/api-info`

Display current API configuration.

Usage:

/api-info

Example:

> /api-info

API Configuration:
  Endpoint: https://app.bike4mind.com/api
  Status: Connected
  Authentication: Valid
  Last Checked: 2 minutes ago

Limits:
  Rate Limit: 100 requests/minute
  Max Context: 200k tokens

Information Shown:

Current API endpoint
Connection status
Authentication state
Rate limits (if available)
Feature availability

When to use:

Verifying configuration
Debugging connection issues
Checking which instance you're connected to

Tool Permission Commands

`/trust <tool>`

Add a tool to the trusted list (won't ask permission again).

Usage:

/trust <tool-name>

Arguments:

tool-name - Name of the tool to trust (e.g., bash_execute, web_search)

Example:

> /trust bash_execute

✓ Tool trusted: bash_execute
This tool will now execute automatically without permission prompts.

To remove: /untrust bash_execute

What happens:

Adds tool to trustedTools in config
Future executions of this tool won't prompt for permission
Permission prompts will show "Automatically allowed (trusted)"

When to use:

You use a tool frequently and trust it
Streamlining workflow (e.g., trusting bash_execute for dev work)
After verifying a tool behaves as expected

Safety Tips:

Only trust tools you fully understand
Be cautious with tools that modify files or run commands
Review what a tool does before trusting it

`/untrust <tool>`

Remove a tool from the trusted list.

Usage:

/untrust <tool-name>

Arguments:

tool-name - Name of the tool to untrust

Example:

> /untrust bash_execute

✓ Tool untrusted: bash_execute
This tool will now require permission before executing.

What happens:

Removes tool from trustedTools in config
Future executions will prompt for permission
Does not affect other tools

When to use:

You want more control over a tool
Security: requiring explicit permission
A tool behaved unexpectedly

`/trusted`

List all currently trusted tools.

Usage:

/trusted

Example:

> /trusted

Trusted Tools (no permission required):
  ✓ bash_execute - Execute shell commands
  ✓ web_search - Search the web
  ✓ weather_info - Get weather information

To remove trust: /untrust <tool-name>

Information Shown:

Tool name
Tool description
How to untrust

When to use:

Reviewing security settings
Checking which tools auto-execute
Auditing trusted tools list

MCP Commands

`/mcp`

Show status of configured and connected MCP (Model Context Protocol) servers.

Usage:

/mcp

Aliases: /mcp:list

Example:

> /mcp

📡 Configured MCP Servers:

• context7 - ✅ Enabled
  Command: npx -y @upstash/context7-mcp

• github - ⏸️  Disabled
  Command: docker run -i ghcr.io/modelcontextprotocol/servers/github
  Env vars: GITHUB_TOKEN

🔌 Connected & Active:

• context7: 2 tools

Total: 2 MCP tools available

What it shows:

Configured servers: All MCP servers defined in your config
Status: Whether each server is enabled (✅) or disabled (⏸️)
Command: The command used to start each server
Environment variables: API keys and config vars (names only, values hidden)
Connected servers: Servers that successfully started and connected
Tool count: Number of tools available from each connected server

Related:

See MCP Integration for complete MCP setup guide
Use b4m mcp add to add new MCP servers (external command)
Use b4m mcp remove to remove servers (external command)
Use b4m mcp enable/disable to toggle servers (external command)

When to use:

Verifying MCP servers loaded successfully
Checking how many tools are available
Debugging MCP connection issues
Confirming server configuration

Troubleshooting:

If a server is enabled but not connected, check debug logs: b4m --verbose
If no tools appear, verify the server command is correct
For Docker servers, ensure Docker is running: docker ps

Utility Commands

`/config`

Display current CLI configuration.

Usage:

/config

Example:

> /config

Configuration (~/.bike4mind/config.json):

Authentication:
  User: john@example.com
  API Endpoint: https://app.bike4mind.com/api
  Token Status: Valid

Trusted Tools:
  - bash_execute
  - web_search

MCP Servers:
  ✓ github (enabled)
  ✗ atlassian (disabled)

Tool API Keys:
  ✓ serper (configured)
  ✓ openweather (configured)

Storage:
  Sessions: ~/.bike4mind/sessions/
  Debug Logs: ~/.bike4mind/debug/

Information Shown:

Authentication status
API configuration
Trusted tools
MCP servers status
Configured API keys
File locations

When to use:

Reviewing your setup
Debugging configuration issues
Checking what's enabled/disabled

`/usage`

Show credit usage and balance information.

Usage:

/usage

Example:

> /usage

Credit Usage:
  Current Balance: 1,250 credits
  Used This Month: 3,420 credits

Recent Usage:
  Today: 145 credits (12 requests)
  Yesterday: 198 credits (15 requests)
  This Week: 847 credits (68 requests)

Average per Request: 28.5 credits
Estimated Monthly Usage: ~4,200 credits

Information Shown:

Current credit balance
Credits used this month
Recent usage breakdown
Average credits per request
Usage projections

When to use:

Checking remaining credits
Monitoring usage patterns
Planning budget
Identifying high-usage activities

`/help`

Show help information and available commands.

Usage:

/help

Example:

> /help

B4M CLI - Interactive AI Assistant

Available Commands:
  Authentication:
    /login              - Authenticate with Bike4Mind
    /logout             - Sign out
    /whoami             - Show current user

  Session Management:
    /sessions           - List saved sessions
    /save <name>        - Save current session

  [... more commands ...]

For detailed documentation: https://docs.bike4mind.com/cli

When to use:

Forgot a command
Quick reference
Finding available commands

`/exit`

Exit the CLI application.

Usage:

/exit

Alternatives:

Press Ctrl+C (sends SIGINT)
Press Ctrl+D (EOF)
Close terminal window

Example:

> /exit

Goodbye! 👋

What happens:

Saves current session state (if applicable)
Closes all connections
Exits process gracefully

When to use:

Ending your CLI session
Switching to a different task

Tip: Consider using /save before exiting to preserve your conversation.

Custom Commands

`/commands`

List all custom commands available in your project.

Usage:

/commands

Example:

> /commands

Custom Commands:
  /review-pr <number>     - Review a pull request
  /deploy-staging         - Deploy to staging environment
  /run-tests              - Run test suite
  /check-deps             - Check for outdated dependencies

Location: .claude/commands/

To create new command: /commands:new <name>

What's shown:

Command name and usage
Brief description
Location of command files

When to use:

Discovering available project commands
Checking command syntax
Finding project-specific workflows

Note: Custom commands are loaded from .claude/commands/ directory in your project.

`/commands:new <name>`

Create a new custom command.

Usage:

/commands:new <command-name>

Arguments:

command-name - Name for the new command (alphanumeric, hyphens)

Example:

> /commands:new review-pr

✓ Created new command: review-pr
Location: .claude/commands/review-pr.md

The file has been created and opened for editing.
Add your command instructions in Markdown format.

Example content:
---
description: Review a pull request
---

Review pull request #{{arg1}}:
1. Check the diff
2. Review for code quality
3. Test the changes
4. Provide feedback

What happens:

Creates .claude/commands/ directory if it doesn't exist
Creates <command-name>.md file
Opens file for editing (if possible)
Command becomes available immediately after saving

File Format:

---
description: Brief description of what this command does
---

Instructions for the AI agent when this command runs.
You can use {{arg1}}, {{arg2}}, etc. for arguments.

When to use:

Creating project-specific workflows
Automating repetitive tasks
Sharing commands with team
Building custom shortcuts

Tips:

Use descriptive names: /deploy-staging, /review-pr
Include clear instructions in the command file
Use arguments for flexibility: {{arg1}}, {{arg2}}
Commit commands to version control for team sharing

Learn more: Custom Commands Guide →

`/commands:reload`

Reload custom commands from disk without restarting the CLI.

Usage:

/commands:reload

Example:

> /commands:reload

✓ Reloaded custom commands
Found 3 custom commands:
  - /review-pr
  - /deploy-staging
  - /run-tests

What happens:

Scans .claude/commands/ directory for command files
Reloads all .md files as custom commands
Updates command list for autocomplete
Makes new/modified commands immediately available

When to use:

After creating new command files manually
After editing existing command files
When commands aren't showing up in /commands list
Testing command file changes without CLI restart

`/project-config`

Display the merged project configuration showing all active settings.

Usage:

/project-config

Example:

> /project-config

Project Configuration:
  Context Files:
    - CLAUDE.md (project, 1.2 KB)
    - ~/.bike4mind/AI.md (global, 0.8 KB)

  Custom Commands: 3 loaded
    - /review-pr (project)
    - /deploy-staging (project)
    - /check-deps (global)

  MCP Servers:
    ✓ github (enabled)
    ✗ atlassian (disabled)

Information Shown:

Active context files (project and global)
Custom commands and their sources
MCP server status
Configuration file locations
Merged settings from all sources

When to use:

Verifying which context files are loaded
Debugging configuration issues
Checking if custom commands are registered
Understanding effective configuration for current project

Sandbox Commands

The sandbox provides OS-level isolation for bash commands. These commands control filesystem restrictions, network filtering, and violation monitoring.

Command	Description
`/sandbox`	Show sandbox status, platform, proxy, and session stats
`/sandbox:enable`	Enable sandbox in auto-allow mode
`/sandbox:disable`	Disable sandbox and stop proxy
`/sandbox:mode <mode>`	Switch between `auto-allow` and `permissions` modes
`/sandbox:trust-domain <domain>`	Add domain(s) to network proxy allowlist
`/sandbox:domains`	List all allowed network domains
`/sandbox:violations [count]`	Show recent violations (default: 20)
`/sandbox:violations:clear`	Clear violation log and reset stats

For detailed sandbox command documentation, see Sandbox Commands →.

Advanced Usage

Command History

Future feature: Navigate previous commands with ↑ and ↓ arrow keys.

# Coming soon
↑  # Previous command
↓  # Next command

Command Chaining

You cannot chain slash commands. Each command must be entered separately:

# ❌ Won't work
> /save my-work /exit

# ✅ Correct
> /save my-work
✓ Session saved
> /exit

Aliases

Several commands have aliases for convenience:

# Available aliases
/quit           → /exit
/new            → /clear
/undo           → /rewind

Example usage:

# These are equivalent
/exit
/quit

# These are equivalent
/clear
/new

# These are equivalent
/rewind
/undo

Command Tips

Case Sensitivity

Commands are case-insensitive:

/login    # ✓ Works
/LOGIN    # ✓ Works
/Login    # ✓ Works

Spacing

Extra spaces are ignored:

/save my-work           # ✓ Works
/save   my-work         # ✓ Works
/save     my-work       # ✓ Works

Incomplete Commands

If you forget arguments, the CLI will prompt you:

> /save
Error: Session name required
Usage: /save <session-name>

Unknown Commands

Unknown commands show an error:

> /unknown
Error: Unknown command: /unknown
Type /help for available commands

Quick Reference​

Authentication Commands​

/login​

/logout​

/whoami​

Session Management Commands​

/sessions​

/save <name>​

/clear​

/rewind​

API Configuration Commands​

/set-api <url>​

/reset-api​

/api-info​

Tool Permission Commands​

/trust <tool>​

/untrust <tool>​

/trusted​

MCP Commands​

/mcp​

Utility Commands​

/config​

/usage​

/help​

/exit​

Custom Commands​

/commands​

/commands:new <name>​

/commands:reload​

/project-config​

Sandbox Commands​

Advanced Usage​

Command History​

Command Chaining​

Aliases​

Command Tips​

Case Sensitivity​

Spacing​

Incomplete Commands​

Unknown Commands​

See Also​

Quick Reference

Authentication Commands

`/login`

`/logout`

`/whoami`

Session Management Commands

`/sessions`

`/save <name>`

`/clear`

`/rewind`

API Configuration Commands

`/set-api <url>`

`/reset-api`

`/api-info`

Tool Permission Commands

`/trust <tool>`

`/untrust <tool>`

`/trusted`

MCP Commands

`/mcp`

Utility Commands

`/config`

`/usage`

`/help`

`/exit`

Custom Commands

`/commands`

`/commands:new <name>`

`/commands:reload`

`/project-config`

Sandbox Commands

Advanced Usage

Command History

Command Chaining

Aliases

Command Tips

Case Sensitivity

Spacing

Incomplete Commands

Unknown Commands

See Also