Built-in Tools

The agent has access to these tools out of the box. It decides when to use them based on your messages.

Tool list

File Operations

Tool	Description
`read_file`	Read the contents of a file
`write_file`	Create or overwrite a file
`edit_file`	Edit specific parts of an existing file
`list_dir`	List directory contents

Terminal & Execution

Tool	Description
`exec`	Execute a shell command
`delegate_task`	Spawn subagents for parallel work or complex subtasks

Web & Search

Tool	Description
`web_search`	Search the web (requires Brave API key)
`web_fetch`	Fetch and extract content from a URL
`session_search`	Search past conversation history

Memory & Organization

Tool	Description
`memory`	Manage persistent memory (MEMORY.md, USER.md)
`todo`	Task list management for complex multi-step work
`clarify`	Ask the user questions when clarification is needed

Skills & Scheduling

Tool	Description
`skills_list`	List available skills
`skill_view`	View a skill’s full content
`skill_manage`	Create, update, or delete skills
`list_cronjobs`	List scheduled tasks
`schedule_cronjob`	Schedule a new recurring or one-time task
`remove_cronjob`	Remove a scheduled task

Messaging

Tool	Description
`message`	Send a message to a connected platform

How tools work

When you send a message, the LLM decides whether it needs to use tools to answer. For example:

“What’s the weather in Tokyo?” → uses web_search or web_fetch
“Create a Python script that…” → uses write_file
“Run my tests” → uses exec
“Summarize this URL” → uses web_fetch
“Remember that my timezone is PST” → uses memory
“Search our past convos about docker” → uses session_search
“Break this down into steps” → uses todo

The agent can chain multiple tools in a single response.

Subagent Delegation

For complex tasks that would flood the context window, the agent can spawn subagents using delegate_task. Each subagent gets:

A fresh conversation (no parent history)
Its own terminal session
A restricted toolset (no recursive delegation, no user interaction)
A focused goal

This is useful for:

Parallel research (research A and B simultaneously)
Debugging sessions that produce lots of output
Code reviews that need to analyze many files

Configuration

Tool-specific settings live under tools in your config. API keys go in ~/.kyber/.env:

# ~/.kyber/.env
KYBER_TOOLS__WEB__SEARCH__API_KEY=YOUR_BRAVE_API_KEY

{
  "tools": {
    "web": {
      "search": {
        "maxResults": 5
      }
    },
    "exec": {
      "timeout": 60,
      "restrictToWorkspace": false
    }
  }
}

See Web Search and Shell Execution for details.

WhatsApp Shell Execution