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

Harnss

Community
Not a standard server
Harness your AI coding agents.

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
Dev Tools #419 of 438
Stars
★ 384 Very popular
Transport
stdio (local process)
Runtime
Prebuilt binary
Credentials
API key / credential required
License
MIT
Last commit
28FMRS · D

Harnss is a feature-rich open-source desktop client that aims to unify management of multiple AI coding agents. It offers excellent tool visualization and integration, but is still early in development with stability concerns.

Strongest · Maintenance 8/20 Weakest · Reliability 4/20

Reliability
4/20
Security and permissions
4/20
Maintenance
8/20
Documentation
6/20
Setup experience
6/20
Why each score
Reliability 4/20
This is an Electron desktop application, not a typical MCP server. The repository contains no MCP server manifest or implementation code, so the MCP handshake or tool listing cannot be verified. The description mentions 'MCP integrations' but no server code is present. Without executable evidence, the reliability score is low.
Security and permissions 4/20
No actual MCP server code is visible in the repository, so credential handling or least privilege cannot be assessed. The app interacts with external AI agents and MCP servers, but no security audit is apparent. It does not meet the 20-point standard, and there are features like a built-in browser and terminal that could be misused, but there is no clear enforcement or confirmation. Potential risks exist but cannot be confirmed from static review alone.
Maintenance 8/20
The repository has an MIT license, is actively developed with recent commits and releases (Stars indicator). However, the README warns of early development and a planned rewrite. There is no clear vulnerability reporting channel or security policy. Maintenance signs are positive but lack a security record and clear governance.
Documentation 6/20
The README is very detailed with screenshots, feature lists, and installation instructions. However, it lacks specific documentation for potential MCP server integration (e.g., how to configure MCP servers, tool parameters, limits). It does not mention costs or troubleshooting. MCP servers are installed, but the documentation is primarily user-oriented, not developer-oriented. Documentation is well structured but lacks depth.
Setup experience 6/20
The README provides instructions to download and run the app, but for MCP server integration, there are no specific configuration examples or API documentation. This is not a traditional MCP server, and configuration appears to be through a GUI. There are no clear steps on how to connect MCP servers from a developer perspective. Installation is straightforward but limited in flexibility.

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 networkControls a browser

Best for

  • Developers who need to manage multiple AI coding agents in parallel
  • Users who want to see what agents actually do with rich visualizations
  • Teams that need integrated terminal, browser, and git workflows in one place

Not for

  • Casual users needing a simple CLI tool
  • Users seeking production-ready stability (currently early development)
  • Users uncomfortable with unsigned binaries

Required permissions

  • Access to local filesystem (open projects, read/write files)
  • Execute local commands (via built-in terminal and agent engines)
  • Network access (for MCP server communication and agent APIs)
  • System notifications (for approval requests, etc.)
  • Microphone (for voice input, optional)

Risks and side effects

  • Early development stage; bugs and instability may occur
  • Pre-built binaries are unsigned, triggering OS warnings
  • Agents can execute arbitrary code; permissions should be carefully controlled
  • Reliance on third-party agents and MCP servers may introduce security risks

Setup

Before you start

Runtime：Prebuilt binary

Download the latest release from the Releases page, open the app, select or create a project folder, then choose an engine (Claude Code, Codex, or an ACP agent) to start working.

Check that it works

Download a release binary from the Releases page, launch the app, open a project folder, and start a session with Claude Code, Codex, or an installed ACP agent; adding an MCP server via the MCP Servers panel and seeing its status and tool count confirms integration works.

Troubleshooting

  1. Ensure Codex CLI is installed and API keys are configured
  2. Check MCP server configuration (transport, URL) for correctness
  3. On macOS, right-click the app and select 'Open' to bypass Gatekeeper on first launch
  4. Consult GitHub Issues or report new issues

Use cases

Run multiple AI coding agent sessions side by side
Visually inspect tool calls and file changes
Manage projects with built-in terminal, browser, and git panels
Connect MCP servers to extend agent capabilities

Overview

Harnss is a cross-platform desktop app that gives you one interface to run, manage, and switch between AI coding agents — Claude Code, Codex, and any ACP-compatible agent — without losing context, sessions, or tool state. It features rich tool visualization, MCP server management, built-in terminal, browser, git, and more.

Similar servers

Context7 80 · B

Upstash's official server providing up-to-date third-party library docs for AI coding assistants

★ 62.9k · Tools 2 Compare with this →
Compare with other MCP servers

Source revision dc1dfd8a33ca Data synced 2026-10-12 Read the FMRS scoring method