M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
D

Dive

Community
Not a standard server
Open-source MCP Host Desktop App that works with any function-calling LLM.

This is not a typical connect-and-use MCP server; read the source repository before using it.

How to install ↓ Compare GitHub source repository ↗
Category
Other
Stars
★ 1.8k Very popular
Transport
stdio (local process) · SSE
Runtime
Node.js · Python · Prebuilt binary
Credentials
Optional API key
License
MIT
Last commit
33FMRS · D

Dive is a feature-rich open-source MCP host desktop app with broad LLM and MCP tool support. Its cross-platform nature and flexibility make it a strong choice for technical users, though some features are unstable and it may have a learning curve for non-technical users.

Strongest · Maintenance 11/20 Weakest · Reliability 2/20

Reliability
2/20
Security and permissions
4/20
Maintenance
11/20
Documentation
10/20
Setup experience
6/20
Why each score
Reliability 2/20
The object is an MCP host desktop app (Electron/Tauri), not an MCP server; no server endpoint, tool list, or init handshake is described or verifiable. No server manifest was cached, and neither CI workflows nor committed tests were available in the static view. Per static calibration, reliability cannot exceed 12 without execution evidence; here there is none, and the underlying premise of a server with declared tools is unconfirmed. Heavily deducted for absent handshake evidence and object misclassification.
Security and permissions 4/20
Red-line-adjacent posture, scored in the 0-4 band. The README discloses built-in "out of the box" tools including Bash (command execution) and File Manager (read/write) with no documented confirmation flow; multiple API keys are stored with an unverifiable storage mechanism; auto-update and OAP cloud integration add external network surfaces; MCP server authentication is self-labeled as unstable. No source-level mitigation, least-privilege design, or data-flow disclosure could be verified. Because dangerous local tools and credential handling are default-enabled and statically unverifiable, security is scored in the red-line band.
Maintenance 11/20
MIT license present; repository not archived; active release history visible (v0.14.0+ with dated update notes); community channels (Discord, Twitter/X) and release download links exist. Deductions: no evidence of CI status, dependency update practice, security-response channel, or issue-closure metrics; 33 open issues without triage evidence; update dates could not be independently verified. Absence of execution/CI evidence prevents a higher score.
Documentation 10/20
The README is detailed on features, platform support, per-OS install, MCP setup options, build, and contribution, and references MCP_SETUP.md and OAPHub integration. Deductions: README/manifest are untrusted evidence; there is no MCP tool parameter reference, auth/credential storage documentation, troubleshooting guide, limits/cost disclosure, or layered documentation; the unstable MCP server authentication is mentioned but not explained. No reviewable evidence beyond prose.
Setup experience 6/20
Per-OS download and install instructions are given (Windows auto-provisions Python/Node; macOS/Linux manual setup; Ubuntu sandbox workaround; Arch AUR). However, this is a host app rather than an MCP server: there is no client configuration example that would let a mainstream MCP client connect to this repository as a server; runtime dependencies are partially manual; no execution evidence exists, so setup may not exceed 15 and is scored low. Docker/CI-based verification is absent.

Static review · not runListed 2026-08-07

Read the FMRS scoring method →

Fit and risk

What it can accessReads local filesWrites / deletes local filesRuns commands or codeUses the network

Best for

  • Developers and technical users who want a desktop interface to various LLMs and MCP tools.
  • Users needing a cross-platform, customizable AI assistant environment.
  • Users managing multiple MCP servers and frequently switching models.

Not for

  • Non-technical users who expect one-click installation with no configuration.
  • Users requiring full stability and official support, as some features are unstable.
  • Users needing headless or CI/CD integration (as a desktop app).

Required permissions

  • Access to the file system via built-in File Manager and Bash tools.
  • Network access via the Fetch tool and MCP server communication.
  • Execute local commands via the Bash tool.
  • Store application data (configuration, chat history).

Risks and side effects

  • MCP server authentication is unstable and may require frequent re-authorization.
  • The Bash tool can execute arbitrary commands, posing a security risk.
  • Misconfiguration may lead to data loss or system issues.
  • Unofficial MCP servers may introduce security threats.

Setup

Before you start

Runtime：Node.js · Python · Prebuilt binary

Download the latest release from GitHub. Windows users can choose the Tauri version (recommended) or Electron. macOS users should download the .dmg of the Electron version. Linux users can choose Tauri or Electron (AppImage). On Windows, Python and Node.js environments will be downloaded automatically after launch; on macOS and Linux, install them manually. See BUILD.md and MCP_SETUP.md for details.

Check that it works

Launch the Dive desktop app and check that the built-in Fetch, File Manager, and Bash tools appear and work in a chat, or connect a managed MCP server via OAPHub.ai one-click deep link to confirm integration.

Troubleshooting

  1. Ensure Python and Node.js (including npx uvx) are installed.
  2. Ubuntu/Debian users may need to add --no-sandbox or adjust sandbox settings when running the AppImage.
  3. Arch users can install via AUR.
  4. If MCP servers fail to connect, check stdio/SSE configuration.
  5. For authentication issues, try re-authorizing.

Things to try

Once connected, you can ask your AI assistant things like:

  • Fetch this web page and summarize it
  • Read the notes file on my desktop and rewrite one paragraph
  • Use the Bash tool to list files in the current directory

Use cases

Interacting with powerful LLMs on the desktop for everyday conversations and tasks.
Integrating multiple AI agents and tools via MCP to automate workflows.
Running Ollama locally while managing multiple MCP servers.

Overview

Dive is an open-source MCP Host desktop application that seamlessly integrates with any LLM that supports function calling. It is cross-platform (Windows, macOS, Linux), supports MCP over stdio and SSE, and includes built-in local tools like Fetch, File Manager, and Bash. It also offers multi-language support, advanced API management, granular tool control, custom instructions, keyboard shortcuts, chat draft saving, and auto-updates.

Similar servers

Telegram MCP Server 70 · B

Telegram MCP server powered by Telethon to let MCP clients read chats, manage groups, and send/modify messages, media, contacts, and settings.

★ 1.8k · Tools 76 Compare with this →
Compare with other MCP servers

Source revision 5fb62b91b55c Data synced 2026-10-12 Read the FMRS scoring method