neuro
ResearchServicesProductsOur WorkBlogAbout UsContactStart a Project

neuro

Engineering growth through AI solutions, bespoke mobile apps, and intelligent automation systems designed to scale.

Quick Links

  • Home
  • Services
  • Research
  • Products
  • Our Work
  • Blog
  • About Us
  • Contact

Services

  • Web Development
  • App Development
  • AI Solutions
  • SaaS Platforms
  • Automation Systems
  • Custom Solutions

Contact

  • support@neuro-corp.com
  • Matara, Sri Lanka

© 2026 Neuro. All rights reserved.

Privacy PolicyTerms & Conditions
Products/Dash/Docs

Getting Started

  • Introduction
  • Installation
  • Quick Start
  • Basic Usage

Configuration

  • AI Providers
  • API Keys
  • Environment Variables
  • Configuration Options

Features

  • Available Tools
  • Project Files
  • Commands and Tool Execution

Help

  • Troubleshooting
  • FAQ

Dash documentation

Available Tools

Every tool registered in Dash, with its inputs and behavior.

Overview#

Dash registers 24 tools with the model. Every tool result is streamed into the conversation, and tool rows are labeled (Read, Edit, List, Search, …) so you can follow what the agent is doing.

File tools#

ToolPurposeKey inputs
read_fileRead a UTF-8 text file from the current workspace.path
write_fileCreate or replace a file, creating folders as needed.path, content
edit_fileReplace one exact text occurrence in a file.path, old_text, new_text
multi_editApply multiple edits to one file atomically — all succeed or none do.path, edits[]
delete_fileDelete, rename or move files/folders.path, action (delete|move|rename), dest
get_file_infoGet size, file type, and modified time for a workspace file.path

Search tools#

ToolPurposeKey inputs
list_filesList files and folders under a workspace directory; optional recursion and glob filtering.path, recursive?, pattern?
find_filesFind workspace files using a glob pattern such as **/*.ts.pattern, path?
search_filesSearch text files for a literal string or regular expression.query, path?, regex?, case_sensitive?
code_searchFind definition and references for a symbol.symbol, kind (definition|references)

Execution tools#

ToolPurposeKey inputs
run_commandRun a shell command in the workspace.command
run_backgroundStart a long-running command, read its output, or kill it.command, action (start|read|kill), pid?
get_diagnosticsRun lint, typecheck, or build and return structured errors.kind (lint|typecheck|build)

Git tools#

ToolPurposeKey inputs
git_toolStructured git operations: status, diff, log, commit, branch, checkout.action, args?

Web tools#

ToolPurposeKey inputs
web_searchSearch the web and return titles, URLs and snippets.query, max_results?
web_fetchFetch a URL and return cleaned markdown, text or HTML.url, format? (markdown|text|html)

Agent coordination tools#

ToolPurposeKey inputs
ask_userAsk the user for clarification or a choice before proceeding.question, options?
todo_writeCreate or update the full task checklist.todos[] (id, title, status)
read_memoryRead the user-approved private memory file across sessions.—
write_memoryPropose a replacement of the memory file — requires user approval to write.content
read_session_activityRead recent conversation activity from this or a previous session.session_id?, limit?
spawn_subagentDelegate a self-contained task to a sub-agent.task, context?
undo_checkpointSave a snapshot before edits or roll back to a checkpoint.action (snapshot|rollback), id?
tool definition (source excerpt)
{
  name: "edit_file",
  description: "Replace one exact text occurrence in a file."
    + " No user approval is needed for edits inside the workspace;"
    + " editing outside the workspace asks the user for approval.",
  parameters: { path: string, old_text: string, new_text: string }
}

Permissions#

  • Workspace writes are trusted: write_file, edit_file, and multi_edit run without approval inside the workspace.
  • Outside the workspace asks first: file operations touching paths outside your project request your approval.
  • Destructive operations always ask: delete_file (delete/rename/move) requires approval.
  • Every shell command is approved by you: run_command prompts for the exact command before it runs.
  • Memory is opt-in: write_memory writes only what you approve.

Reference

See Commands and Tool Execution for slash commands and the approval flow in detail.

On this page

  • Overview
  • File tools
  • Search tools
  • Execution tools
  • Git tools
  • Web tools
  • Agent coordination tools
  • Permissions
PreviousConfiguration Options
NextProject Files

© 2026 Neuro. All rights reserved.