OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
MCP & Tooling

Hound

Hound is a self-hosted Python MCP server that exposes web search, page fetching and crawling, screenshot capture, and PDF content extraction to compatible AI clients. Its documented fetch flow can use HTTP first and escalate to a browser when blocked; PDF handling includes OCR for scanned documents.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Hound — OpenViST project profile in MCP & Tooling.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Hound when an MCP-compatible client needs web retrieval tools available within an agent workflow, especially when the workflow combines search with fetching or crawling, or needs to read PDFs.

Who it is for

  • People configuring an MCP-compatible AI assistant to retrieve web content during agent workflows.
  • Developers who want to run web research tools locally or deploy an HTTP MCP service with Docker.
  • Teams building agent workflows that need page crawling or PDF text extraction.

Documented uses

  • Search the web and return ranked results to an AI agent.
  • Fetch a URL, extract readable page content, or capture a screenshot.
  • Crawl pages within a domain and return page content as Markdown.
  • Extract text from PDFs, including scanned documents through OCR.
Before you choose

Requirements & limits

The repository documents a pip installation, a Docker deployment, and MIT licensing. Its README describes keyless search and separately supports adding keys for named search providers.

  • The primary interface is an MCP server, so an MCP-compatible client is needed to use its tools.
  • The full setup installs Chromium through Playwright; the documented lean package omits browser dependencies and runs in HTTP-only mode with reduced capability.
  • Docker Compose provides an HTTP MCP endpoint for clients that support HTTP transport.
  • Search can use documented keyless local engines or configured provider keys.
  • The project documents bot-protection handling, but retrieval from a protected site is not guaranteed.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Install Hound for a local MCP client

Install the full browser-enabled Hound package before connecting an MCP client.

Before you start

  • Python and pip
  • Playwright and Chromium installation support

Steps

  1. Install the package with the command below.

    Documented command
    pip install hound-mcp[all]

  2. Install Chromium with playwright install chromium.

  3. Point an MCP client at the hound command.

Expected outcome

The MCP client is configured to connect to the locally installed Hound command.

Guide limitations

  • This guide covers installation and connection configuration, not a verified client workflow.
  • Browser-based retrieval requires the documented browser setup.
Documentation 1 ↗
Source-based practical guide

Start Hound as an HTTP MCP service with Docker Compose

Launch the documented Docker Compose deployment and identify its MCP endpoint.

Before you start

  • Docker and Docker Compose
  • A host address reachable by the MCP client

Steps

  1. Clone the Hound repository.

  2. Change into the master-fetch directory.

  3. Run docker-compose up -d.

  4. Configure an HTTP-capable MCP client to use http://:8765/mcp, or localhost when connecting on the same machine.

Expected outcome

Hound is configured as an HTTP MCP server at the documented endpoint.

Guide limitations

  • The README identifies HTTP-capable clients as a requirement for this connection method.
  • The documented steps do not establish that every deployment environment or client will work identically.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
12 October 2026
Original website title
dondai1234/master-fetch · Official homepage title · retained in its source language
How we review sourcesSuggest a correction