> ## Documentation Index
> Fetch the complete documentation index at: https://docs.gcaplabs.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Tools & Toolsets
> 60+ built-in tools across 10+ categories — web search, terminal, browser, image generation, code execution, delegation, and more.
# Tools & Toolsets
Tools are functions that extend the agent's capabilities, organized into logical toolsets that can be enabled or disabled per platform and per specialist. Headmaster ships with a broad built-in tool registry — over 60 tools across 10+ categories.
***
## Available tool categories
| Category | Tools | Description |
| ------------------------- | ----------------------------------------------------------------------------------------- | ------------------------------------------------------------- |
| **Web** | `web_search`, `web_extract` | Search the web and extract page content as markdown |
| **X (Twitter) search** | `x_search` | Search X/Twitter — gated on xAI credentials, off by default |
| **Terminal & files** | `terminal`, `process`, `read_file`, `patch`, `write_file`, `search_files` | Execute commands, read/write/edit files, search file contents |
| **Browser** | `browser_navigate`, `browser_snapshot`, `browser_click`, `browser_type`, `browser_vision` | Interactive browser automation — text and vision based |
| **Media** | `vision_analyze`, `image_generate`, `text_to_speech` | Analyze images, generate images, convert text to speech audio |
| **Agent orchestration** | `todo`, `clarify`, `execute_code`, `delegate_task` | Planning, clarification, code execution, subagent delegation |
| **Memory & recall** | `memory`, `session_search` | Persistent memory and cross-session search |
| **Automation & delivery** | `cronjob`, `send_message` | Scheduled tasks and outbound messaging to channels |
| **Integrations** | Home Assistant, MCP server tools, Spotify, Discord | Smart home, external tools, music, chat platforms |
| **Delegation** | `delegate_task` | Spawn isolated subagents for parallel workstreams |
| **Advanced reasoning** | `mixture_of_agents` | Route a hard problem through multiple LLMs collaboratively |
***
## Enabling and disabling tools
Go to **Settings → Specialists** and pick a specialist.
Go to the **Tools** section and check/uncheck toolsets.
A toolset toggled off is not loaded — the agent won't see or try to use those tools.
Tools can be configured per platform (desktop, Telegram, Discord, etc.). This lets you give the desktop agent full tool access while restricting the Telegram agent to safe tools only.
Go to **Settings → Headmaster's Library → Advanced → Toolsets**.
Pick a platform and toggle toolsets on or off for that platform.
### Platform presets
| Preset | Tools included |
| ------------ | ---------------------------------------------------------------------------------- |
| **Desktop** | All tools — full power |
| **Telegram** | Web, terminal, file, memory, skills, cron, messaging (no destructive tools) |
| **Discord** | Web, terminal, file, memory, skills, cron, messaging |
| **CLI** | All tools — same as desktop |
| **Safe** | Web, memory, skills, clarify, todo (no terminal, no file write, no code execution) |
***
## Using tools from chat
You don't need to explicitly call tools — the agent picks the right tool based on your request:
```
Search the web for the latest Python release notes and summarize the key changes.
```
→ Agent uses `web_search` and `web_extract`.
```
Run `git status` in my project folder.
```
→ Agent uses `terminal`.
```
Take a screenshot of https://example.com.
```
→ Agent uses `browser_navigate` and `browser_vision`.
```
Generate an image of a mountain landscape at sunset.
```
→ Agent uses `image_generate`.
### Tool call display
Every tool call appears in the chat as an expandable block:
```
🔧 web_search("latest Python release notes")
→ 5 results found
→ Extracting content from python.org...
→ 3,200 characters extracted
```
Click to expand and see the full tool input and output. This is your audit trail — every tool call is logged.
***
## execute\_code — programmatic tool calling
The `execute_code` tool lets the agent run Python scripts that call other Headmaster tools programmatically. This collapses multi-step pipelines into a single inference call.
Instead of the agent calling `web_search` 5 times in sequence, it can write a Python script that calls `web_search` 5 times in a loop, processes the results, and returns a single summary — faster, cheaper, and more reliable.
**Example:**
```
Search for the top 5 Python testing frameworks, extract their docs, and compare their features in a table.
```
The agent writes a Python script that calls `web_search` for each framework, `web_extract` on each result, parses the content, builds a comparison table — all in one tool call instead of 10+.
***
## mixture\_of\_agents — collaborative reasoning
The `mixture_of_agents` tool routes a hard problem through multiple frontier LLMs collaboratively. It makes 5 API calls (4 reference models + 1 aggregator) with maximum reasoning effort.
**Use it for:**
* Complex math or algorithms.
* Multi-step analytical reasoning.
* Problems that benefit from diverse perspectives.
The agent decides when to use it — you don't need to explicitly request it. It's used sparingly because it's expensive (5 API calls per use).
***
## Bundled tools
Some tools — web search, image generation, text-to-speech, and browser automation — can be enabled with a single bundled configuration instead of separate API keys for each service.
When the bundle is configured:
* **Web search** works out of the box — no Firecrawl or SerpAPI key needed.
* **Image generation** works — no separate FAL or OpenAI images key needed.
* **Text-to-speech** works — no OpenAI TTS key needed.
* **Browser automation** works — no separate browser service needed.
***
## Tool configuration reference
| Tool | Setting location | What it controls |
| ------------------ | --------------------------- | --------------------------------------------- |
| Web search | Settings → Advanced → Tools | Enable/disable, choose backend |
| Image generation | Settings → Advanced → Tools | Enable/disable, choose model |
| Text-to-speech | Settings → Advanced → Tools | Enable/disable, choose voice and provider |
| Browser automation | Settings → Advanced → Tools | Enable/disable, choose backend |
| Terminal | Settings → Advanced → Tools | Enable/disable, choose backend |
| X (Twitter) search | Settings → Advanced → Tools | Enable/disable (requires xAI credentials) |
| Home Assistant | Settings → Advanced → Tools | Enable/disable (requires HA URL and token) |
| Spotify | Settings → Advanced → Tools | Enable/disable (requires Spotify credentials) |
| Discord | Settings → Advanced → Tools | Enable/disable (requires bot token) |
***
Local, Docker, SSH, Singularity, Modal, and Daytona.
Full browser automation details.
Full image generation setup guide.
Extend tools with MCP servers.