Skip to main content

​
Overview

Below is the full description of the Built-in Tools included in Devic, organized by functional purpose. Built-in-tools description

​
Read Emails

Reads and manages email messages.
  • Available tools:
    • get_unread_emails, get_last_emails, mark_email_as_read, get_email_attachment, among others.
Read-emails description

​
Send Email

Allows automated email sending.
  • Tool: send_basic_email
  • Description: Sends an email to the specified address and can listen for a reply.
  • Recommended use:
    • Sending personalized emails.
    • Automated notifications or reports.
    • Internal alerts or agent-to-agent communication.

​
Code Execution Tools

Executes code in different languages directly from Devic (currently Python).
  • Tool: execute_python_code
  • Description: Executes Python code and returns the result or an error.
  • Recommended use:
    • Data processing or logical validation.

​
Terminal Sandbox

Gives the assistant or agent an isolated Linux machine where it can run code and shell commands. This is the tool group to add when a skill or a task requires executing something — installing a package, running a Node or Python script, building a file — rather than just reasoning about it.
  • Available tools:
    • create_terminal_sandbox, run_terminal_command, stop_terminal_sandbox, get_sandbox_status, extend_sandbox_timeout
    • Files: write_sandbox_file, read_sandbox_file, list_sandbox_files, create_sandbox_directory
    • Transfers: download_sandbox_file (returns a shareable URL), upload_file_to_sandbox
    • Serving: get_sandbox_public_url
  • Runtimes: Node.js and Python, with full shell access.
  • Recommended use:
    • Generating files (documents, presentations, spreadsheets) with code.
    • Running scripts a skill defines.
    • Anything that needs real execution and a filesystem.
Sandboxes are ephemeral: files are lost when the sandbox stops unless snapshots are enabled. To hand a generated file to the user or to another step, get a URL with download_sandbox_file.

​
Browser Sandbox

Browses the web with a real headless Chromium inside a sandbox, when reading a page is not enough and the assistant needs to interact with it.
  • Available tools: create_browser_session, browser_navigate, browser_observe, browser_action, browser_get_element, close_browser_session
  • Recommended use:
    • Sites that require clicking, typing or scrolling to reach the content.
    • Visual checks through screenshots.

​
Search Knowledge

Searches for relevant information within internal knowledge bases.
  • Tool: retrieve_from_rag_in_file
This tool only works if at least one document has been added.
Explores the knowledge base like a filesystem instead of only searching it semantically. Required for skills: it is what lets a model open a skill’s instructions on demand.
  • Available tools: list_knowledge_directory, read_knowledge_document, grep_knowledge, search_knowledge
  • Recommended use:
    • Attaching skills (without this group their full text is injected into the system prompt on every run).
    • Browsing folders and reading specific documents by id.
    • Searching exact text patterns across the knowledge base.

​
Update Knowledge

Updates the knowledge base of agents and MCPs through write or synchronization operations.
  • Tool: update_memory
  • Description: Each agent maintains its own memory state, which can be updated by adding new text fragments (e.g., descriptions, notes, or results). This update only affects the agent’s internal memory — it does not modify source documents.
  • Recommended use:
    • Add or update contextual knowledge for the agent.
    • Synchronize information from documents or external APIs.
    • Allow the agent to “learn” new data incrementally while keeping original sources unchanged.
Allows searching for up-to-date information on the web.
  • Available tools:
    • search_web
    • web_extract

​
Databases

Creates, queries, updates, and deletes structured information in internal databases.
  • Tools:
    • get_tables, create_table, filter_rows, update_row, vector_search_table_rows, among others.

​
Files and Directories

Manages documents and folders within the Devic workspace.
  • Available tools:
    • create_folder, get_folder_content, move_document_to_folder, navigate_to_folder, among others.

​
OCR

Extracts text from images or scanned documents.
  • Available tools:
    • ocr_document
    • query_document

​
PDF from HTML

Converts HTML content into visually formatted PDF documents.
  • Tool: generate_pdf_from_html
  • Recommended use:
    • Creating corporate reports or styled templates.

​
PDF from Markdown

Generates PDF files from Markdown text.
  • Tool: generate_pdf_from_markdown

​
Image Generation

Generates images from a text prompt.
  • Tool: generate_image
  • Recommended use:
    • Illustrations or visual assets for a document or a presentation.

​
Image Viewer (vision)

Loads an image into the model’s visual context so it can actually look at it.
  • Tool: view_image
  • Note: this group is added automatically for models that support vision — you do not need to assign it.

​
Chart Images

Generates visual charts from structured data.
  • Tool: generate_chart_image
  • Description: Creates bar, line, or area charts.
  • Recommended use:
    • Data visualization or dynamic reports.

​
Spreadsheet Tools

Allows reading and manipulating spreadsheets (Excel or CSV).
  • Tools:
    • get_sheets, get_sheet_content

​
Text File Templates

Generates documents from templates (Word, Excel, PDF).
  • Available tools:
    • fill_text_template, get_text_templates, merge_word_documents

​
SharePoint API

Integrates Devic with Microsoft SharePoint.
  • Tools:
    • check_active_site, get_document_details, get_excel_content, update_cell

​
Short and Medium Term Memory

Allows agents to access the context of previous interactions.
  • Tools:
    • get_agent_threads_summary, query_agent_thread_details

​
Group IDs for the API and the CLI

In the dashboard you pick these groups from a list. Over the API or the CLI you reference them by UID, inside availableToolsGroupsUids — the same flat array that also holds your own MCP servers’ ids:
For an agent the field lives inside assistantSpecialization; for an assistant it sits at the root. Adding a group grants every tool it contains — narrow it down with enabledTools if you only want some.
This table lists the groups you can assign today. Availability can differ by account and by plan, so if a group described above is not offered in your dashboard, it is not enabled for you — check there before wiring its UID.

​
Next Steps

Agents

Learn how to create and configure autonomous agents capable of executing complex processes and making context-aware decisions.

Assistants

Discover how assistants provide rich conversational experiences and integrate with your tools and business data.

Databases

Explore how to create, manage, and connect relational or vector databases to enhance search and reasoning capabilities.

MCPs

Dive into the Model Context Protocol (MCP) standard and learn to connect your internal APIs and services as intelligent tools.