Skip to content
MCP server

chargehand

Ask a code question and get claims that each cite a file at a pinned commit

Runs locallyStatus and history
Install
Pricing
Open source
Platforms
CLI, self-hosted
Runs
Locally
Version
0.19.29
Listing updated
today

Similar MCP servers

About chargehand

chargehand is an open-source MCP server for developer tools. It runs locally on your machine and is started by your AI client.

Coding agents usually answer in prose; chargehand returns something a program can check. It runs a read-only Claude Code or OpenCode session on a clone of your repository at a pinned commit and returns typed JSON with claims, open questions, a confidence score, cost and the prompts used. Each citation, whether a file and line range, a commit, a diff range or a session message, is verified before the claim leaves its worker. A worker gets one repair turn, and claims that still fail are moved to open questions.

The README is open about the limits. Resolving a citation shows the cited thing exists, not that it supports the claim; that check is planned. Splitting a question into subtasks cost about 1.53 times a plain session for a small score gain, so the default presets do not split. The project is pre-1.0 and lists one user, its author. It does not call model providers itself, host a service or keep a memory store.

Beyond the MCP server there is an HTTP API, a CLI, a browser console, a code preset that writes a tested branch or draft pull request in a sandbox, and Prompt CI that gates prompt changes on evals. It needs the .NET 10 SDK and Claude Code with a token.

Install

Run in a terminal

claude mcp add chargehand --env CLAUDE_CODE_OAUTH_TOKEN='<CLAUDE_CODE_OAUTH_TOKEN>' --env ANTHROPIC_API_KEY='<ANTHROPIC_API_KEY>' -- dnx Chargehand --yes

Add to claude_desktop_config.json

{
  "mcpServers": {
    "chargehand": {
      "command": "dnx",
      "args": [
        "Chargehand",
        "--yes"
      ],
      "env": {
        "CLAUDE_CODE_OAUTH_TOKEN": "<CLAUDE_CODE_OAUTH_TOKEN>",
        "ANTHROPIC_API_KEY": "<ANTHROPIC_API_KEY>"
      }
    }
  }
}

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "chargehand": {
      "command": "dnx",
      "args": [
        "Chargehand",
        "--yes"
      ],
      "env": {
        "CLAUDE_CODE_OAUTH_TOKEN": "<CLAUDE_CODE_OAUTH_TOKEN>",
        "ANTHROPIC_API_KEY": "<ANTHROPIC_API_KEY>"
      }
    }
  }
}

Add to .vscode/mcp.json

{
  "servers": {
    "chargehand": {
      "type": "stdio",
      "command": "dnx",
      "args": [
        "Chargehand",
        "--yes"
      ],
      "env": {
        "CLAUDE_CODE_OAUTH_TOKEN": "<CLAUDE_CODE_OAUTH_TOKEN>",
        "ANTHROPIC_API_KEY": "<ANTHROPIC_API_KEY>"
      }
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "chargehand": {
      "command": "dnx",
      "args": [
        "Chargehand",
        "--yes"
      ],
      "env": {
        "CLAUDE_CODE_OAUTH_TOKEN": "<CLAUDE_CODE_OAUTH_TOKEN>",
        "ANTHROPIC_API_KEY": "<ANTHROPIC_API_KEY>"
      }
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "chargehand": {
      "command": "dnx",
      "args": [
        "Chargehand",
        "--yes"
      ],
      "env": {
        "CLAUDE_CODE_OAUTH_TOKEN": "<CLAUDE_CODE_OAUTH_TOKEN>",
        "ANTHROPIC_API_KEY": "<ANTHROPIC_API_KEY>"
      }
    }
  }
}

Using several servers? Build one config file with the MCP config generator.

Configuration

  • CHARGEHAND_RUNTIMEWorker runtime, when both or neither of claude and opencode are on PATH.
  • CHARGEHAND_PROFILEA profile/v1 file for budgets, prices, memory or a pinned runtime version.
  • CLAUDE_CODE_OAUTH_TOKENClaude Code subscription token from claude setup-token. Set this or ANTHROPIC_API_KEY.
  • ANTHROPIC_API_KEYAnthropic API key for Claude Code workers. Set this or CLAUDE_CODE_OAUTH_TOKEN.

What it does

Checked citations
Every claim must cite something that exists at the pinned commit.
Typed results
Returns result/v1 JSON with claims, open questions, confidence and cost.
Presets
default, cheap, thorough, strict, draft and review set tools, budgets and splitting.
Multiple interfaces
Use it as an MCP server, an HTTP API or a CLI.
Writing mode
A code preset produces a tested branch or draft pull request in a sandbox.
Prompt CI
Gates prompt and preset changes on paired evals.
Run log
Records tokens, cache hits, cost and the prompt chain of every call.

Good for

  • Find where a behaviour is decided in a codebase, with file references
  • Give a program a verifiable answer about a repository
  • Budget and cap what a code question may cost
  • Gate prompt changes on evals before shipping them

Pricing

Open source

chargehand FAQ

What does it need to run?

The .NET 10 SDK, Claude Code signed in, and a token from claude setup-token or an Anthropic API key.

Which agents do the work?

Claude Code or OpenCode sessions; chargehand never calls a model provider itself.

Does it check that a claim is true?

Not yet. It checks that each citation exists at the pinned commit; verifying that the cited text supports the claim is planned.

Is it stable?

The README marks it as before 1.0 and advises pinning an exact version.

Is chargehand free?

Yes. chargehand is open source, so you can use and self-host it for free.

How do I add chargehand to Claude Code?

Run this command in your terminal: claude mcp add chargehand --env CLAUDE_CODE_OAUTH_TOKEN='<CLAUDE_CODE_OAUTH_TOKEN>' --env ANTHROPIC_API_KEY='<ANTHROPIC_API_KEY>' -- dnx Chargehand --yes

What are the best alternatives to chargehand?

The closest alternatives to chargehand are REA, Context7, Chrome DevTools MCP, Codebase Memory and GitHub.

Where this comes from

  • Description: written from the project's README.
  • Install steps: from the official MCP Registry entry.

How AIPoweredBot checks servers