Lobstack is a managed cloud platform for deploying autonomous AI agents. Each agent runs on a dedicated virtual machine with persistent memory, 700+ integrations, and goes live in under 90 seconds with zero DevOps.

How long does it take to deploy an AI agent?

You can deploy an AI agent on Lobstack in under 90 seconds. Describe what your agent should do, pick a model, and Lobstack handles all the infrastructure — no Docker, Kubernetes, or server setup needed.

What AI models does Lobstack support?

Lobstack supports 16 AI models including Claude Opus and Sonnet from Anthropic, GPT-5 and GPT-4o from OpenAI, Gemini 2.5 Pro from Google, plus DeepSeek, Mistral, and Grok. You can hot-swap models without losing agent memory or configuration.

Is Lobstack secure for enterprise use?

Yes. Every agent runs in a gVisor-sandboxed VM with AES-256 encryption at rest, TLS 1.3 in transit, HashiCorp Vault for secrets, row-level security on all tables, and full audit logging. Lobstack is SOC 2 aligned.

Docs/Tools/Built-in Tools

Reference

Tools

Your agent comes with 9 built-in tools for terminal access, file operations, web browsing, and HTTP requests. Skills add 100+ more integration tools.

💡

Tools are real actions

When your agent uses a tool, it executes a real action on its dedicated server — running a shell command, writing a file, or making an HTTP request. These are not simulated responses.

Built-in Tools#

These 9 tools are always available on every agent, regardless of plan or skill configuration. They run directly on your agent's dedicated VM.

Terminal#

Execute shell commands on your agent's Linux server. Full root access with support for any CLI tool installed on the VM.

Parameter	Type	Description
command	string (required)	The shell command to execute
timeout	integer	Max execution time in seconds (default: 30, max: 120)
working_directory	string	Directory to run the command in (default: /root)

File Read#

Read the contents of any file on the server. Supports text files with a default limit of 50KB.

Parameter	Type	Description
file_path	string (required)	Absolute path to the file
max_bytes	integer	Maximum bytes to read (default: 50000)

File Write#

Write or append content to a file. Automatically creates intermediate directories if they don't exist.

Parameter	Type	Description
file_path	string (required)	Absolute path to the file
content	string (required)	Content to write
append	boolean	Append to file instead of overwriting (default: false)

File List#

List the contents of a directory. Supports recursive listing with a default limit of 200 entries.

Parameter	Type	Description
directory_path	string (required)	Absolute path to the directory
recursive	boolean	List recursively (default: false)
max_entries	integer	Maximum entries to return (default: 200)

Web Browse#

Browse a webpage using a headless Chromium browser with full JavaScript rendering. Extracts text content and links from the page.

Parameter	Type	Description
url	string (required)	The URL to browse
extract_links	boolean	Also extract all links from the page (default: false)

Web Scrape#

Extract structured data from a webpage using CSS selectors. Useful for pulling specific elements like prices, titles, or tables.

Parameter	Type	Description
url	string (required)	The URL to scrape
selector	string (required)	CSS selector to target elements
attribute	string	Element attribute to extract (default: text content)

Web Screenshot#

Take a screenshot of a webpage. Captures the full page or just the visible viewport and saves it to the workspace.

Parameter	Type	Description
url	string (required)	The URL to screenshot
full_page	boolean	Capture full page or just viewport (default: true)
output_path	string	Where to save the screenshot

Web Click & Extract#

Perform multi-step interactions on a webpage — click buttons, type into fields, select options, scroll, and wait for elements. Useful for interacting with dynamic web apps.

Parameter	Type	Description
url	string (required)	The URL to interact with
actions	array (required)	List of actions: click, type, select, wait, scroll
extract_after	string	CSS selector to extract content after actions complete

HTTP Request#

Make arbitrary HTTP requests to external APIs. Supports all methods, custom headers, and request bodies.

Parameter	Type	Description
url	string (required)	The URL to request (must be external, not localhost)
method	string	HTTP method: GET, POST, PUT, DELETE, PATCH (default: GET)
headers	object	Custom request headers
body	string	Request body (for POST/PUT/PATCH)

Skill-Integrated Tools#

When you enable a skill and provide API credentials, your agent gets additional tools specific to that integration. These are real API-backed tools — your agent makes actual calls to Discord, GitHub, Gmail, and other services.

Skill	Tools Provided
Discord	send_message, read_messages, list_channels, list_guilds
Slack	send_message, read_messages, list_channels
Telegram	send_message, read_updates, set_webhook
Twitter/X	read_timeline, post_tweet, search, get_mentions
WhatsApp	send_message, read_messages

✅

100+ integration tools

The full list includes 100+ tools across social media, developer platforms, databases, CRMs, email providers, search engines, AI models, analytics, finance, and more. Enable any skill from the Skills panel to unlock its tools.

Tool Execution#

When your agent decides to use a tool, here's what happens:

The AI model returns a tool call with the tool name and parameters
The bridge executes the tool on the agent's VM
The result is sent back to the AI model as tool output
The model uses the result to formulate its response
This loop can repeat up to 25 times per message for complex multi-step tasks

Execution Constraints#

Constraint	Value
Max tool rounds per message	25
Terminal command timeout	120 seconds
File read limit	50 KB (default)
Directory listing limit	200 entries (default)
Tool result truncation	20 KB max per result

TOOLS.md Configuration#

The TOOLS.md config file lets you customize how your agent uses its tools. Use it to set tool priorities, define error handling behavior, and document interaction patterns.

TOOLS.md (example)

# TOOLS — Tool Usage Patterns

## Usage Priority
1. Check memory for past context and solutions
2. Search available knowledge bases and documentation
3. Use terminal for system operations and diagnostics
4. Use HTTP requests for external API calls
5. Use web browsing for research and data gathering

## Error Handling
1. Tool fails -> try alternative approach
2. Data not found -> ask for clarification
3. Permission denied -> explain what's needed
4. System down -> set expectations and follow up

## Best Practices
- Prefer file operations over terminal for reading/writing files
- Use web_browse for dynamic pages, http_request for APIs
- Always validate terminal command output before acting on it
- Save important results to workspace files for later reference

Multi-Provider Support#

Tools work identically across all AI providers. Whether your agent runs on Claude, GPT-4o, Gemini, Grok, or any other supported model, the same tools are available with the same capabilities. The bridge automatically converts tool schemas to each provider's native format.

Provider	Tool Format
Anthropic (Claude)	Native tool_use format with multi-turn support
OpenAI (GPT)	Function-calling format with tool_choice
Google (Gemini)	OpenAI-compatible function calling
xAI (Grok)	OpenAI-compatible function calling

💡

Switch models anytime

Changing your AI model doesn't affect tool availability. All tools work the same regardless of which provider powers your agent. See Supported Models.

Memory System Skills Overview