Codequiry MCP Server

Give your AI assistant the ability to check source code for plagiarism, peer similarity, and AI-generated code, using Codequiry.

Works with any MCP client: Claude.ai, Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, Zed.

You: Two of these submissions look suspiciously alike. Zip them up and check.
Claude: [codequiry_quick_check] ... [codequiry_wait_for_check] ... [codequiry_get_check_summary]
        submission_04 and submission_11 share 87% of their code, across 6 files.
        Want me to pull up the matching lines side by side?

Two ways to connect

Sign in, nothing to install. Codequiry also runs this server at https://codequiry.com/mcp. Add that address as a connector and your assistant opens a Codequiry sign-in; press Allow once and it is connected to your account. No key to copy, and you can disconnect it any time from API keys in the dashboard. Because it runs on codequiry.com it cannot see your disk: give it a public link to a .zip (a GitHub "Download ZIP" link works) or let the assistant pass the archive content.

| Client | Where the URL goes | | --- | --- | | Claude.ai, Claude Desktop | Settings, Connectors, Add custom connector, URL https://codequiry.com/mcp | | Claude Code | claude mcp add --transport http codequiry https://codequiry.com/mcp, then /mcp and Authenticate | | Cursor | ~/.cursor/mcp.json: {"mcpServers":{"codequiry":{"url":"https://codequiry.com/mcp"}}} | | VS Code | code --add-mcp '{"name":"codequiry","type":"http","url":"https://codequiry.com/mcp"}' | | Windsurf | mcp_config.json: {"mcpServers":{"codequiry":{"serverUrl":"https://codequiry.com/mcp"}}} |

Run it locally with an API key. This package. It runs on your machine, so it can zip and upload folders straight from your disk. The rest of this README is about that.

Setup

You need a Codequiry API key. Create an account, then copy the key from API Keys in the dashboard.

Claude Code

claude mcp add codequiry --env CODEQUIRY_API_KEY=your_key_here -- npx -y codequiry-mcp

Add --scope user to make it available in every project. Check with claude mcp list.

Claude Desktop

Edit claude_desktop_config.json (Settings, Developer, Edit Config):

{
  "mcpServers": {
    "codequiry": {
      "command": "npx",
      "args": ["-y", "codequiry-mcp"],
      "env": {
        "CODEQUIRY_API_KEY": "your_key_here"
      }
    }
  }
}

Restart Claude Desktop.

Cursor

Edit ~/.cursor/mcp.json, or .cursor/mcp.json in a project, with the same block as Claude Desktop.

VS Code

code --add-mcp '{"name":"codequiry","command":"npx","args":["-y","codequiry-mcp"],"env":{"CODEQUIRY_API_KEY":"your_key_here"}}'

Windsurf

Edit ~/.codeium/windsurf/mcp_config.json with the same block as Claude Desktop.

Zed

In settings.json:

{
  "context_servers": {
    "codequiry": {
      "command": {
        "path": "npx",
        "args": ["-y", "codequiry-mcp"],
        "env": { "CODEQUIRY_API_KEY": "your_key_here" }
      }
    }
  }
}

Windows

If a client reports that it cannot start npx, go through cmd:

"command": "cmd",
"args": ["/c", "npx", "-y", "codequiry-mcp"]

Node 18 or newer is required everywhere.

Tools

Every operation the REST API offers, as one tool each, plus wait_for_check. The signed-in server offers the same 26 tools; its upload tools take {url} or {base64} archives instead of file paths.

Run a check

| Tool | What it does | Balance | | --- | --- | --- | | codequiry_quick_check | Create, upload and start a check in one call. The default path. | Spends 1 | | codequiry_wait_for_check | Poll until the check finishes, with backoff. | Free | | codequiry_create_check | Create an empty check, optionally with a completion webhook. | Free | | codequiry_upload_submission | Upload one ZIP into an existing check. | Free | | codequiry_upload_submissions | Upload up to 50 ZIPs into an existing check in one request. | Free | | codequiry_start_check | Start analysis on a check that has submissions. | Spends 1 |

Read results

| Tool | What it does | | --- | --- | | codequiry_get_check_summary | Headline numbers: average, maximum, AI scores, how many are flagged. | | codequiry_get_overview | Per-submission similarity scores for a finished check. | | codequiry_get_results | Matched files and line ranges for one submission. | | codequiry_get_ai_results | AI-generated-code detection scores, per submission and per file. | | codequiry_export_results | One row per submission, as JSON or CSV text. | | codequiry_view_submission | Everything the dashboard's results page shows for one submission: files, matches, web sources, and the score rail. | | codequiry_get_submission_file | The text of one file inside a submission. | | codequiry_get_match_counterpart | The code on the other side of a match, with line ranges on both sides. | | codequiry_get_matched_web_source | The stored copy of an external page a submission matched. |

Manage checks

| Tool | What it does | Balance | | --- | --- | --- | | codequiry_list_checks | Past checks, newest first, optionally with stats beside each. | Free | | codequiry_get_check | One check with its settings, status and submissions. | Free | | codequiry_get_check_status | Status and progress of one check. | Free | | codequiry_update_check | Rename a check or change its language. | Free | | codequiry_delete_check | Delete a check and everything in it. | Deletes | | codequiry_delete_submission | Delete one submission from a check. | Deletes |

Account and reference

| Tool | What it does | | --- | --- | | codequiry_account | The account behind the key and how many checks remain. | | codequiry_validate_key | Confirm the key is accepted. The first thing to try when setup fails. | | codequiry_service_status | Engine health and queue depth. | | codequiry_list_languages | Language ids for language. 999 is auto-detect. | | codequiry_list_test_types | Engine ids for test_type. |

Each submission is one ZIP archive, up to 10 MB. To compare several people's work, pass several ZIPs to one check.

Configuration

| Variable | Default | Purpose | | --- | --- | --- | | CODEQUIRY_API_KEY | required | Your 64-character key from the dashboard. | | CODEQUIRY_MCP_READONLY | unset | Set to 1 to hide every tool that writes, leaving the eighteen that only read. The assistant can still go through every past result but cannot create, upload, start, rename or delete anything. | | CODEQUIRY_API_BASE | https://codequiry.com/api/v1 | Override the API base URL. |

Cost and safety

Starting a check consumes one check from your plan balance. Only codequiry_quick_check and codequiry_start_check do that. Those two, and the two delete tools, are annotated destructive, so well-behaved clients ask before running them. Creating a check, uploading to it and renaming it are writes but cost nothing and are undoable, so they are not flagged: a confirmation prompt on a free action just teaches people to click through the one that matters.

If you want a hard guarantee rather than a prompt, run with CODEQUIRY_MCP_READONLY=1, or connect the signed-in server with the read scope alone.

The API allows 60 requests per minute per IP, shared across all endpoints. The server retries 429s using the server's own Retry-After. Use codequiry_wait_for_check rather than polling in a loop.

Development

npm install
npm test

npm test boots the server over stdio as a real client and runs 103 checks against a scripted mock API, which is how the tools that spend or delete get exercised without spending or deleting anything. npm run test:live calls every read-only tool against the real API using CODEQUIRY_API_KEY or the key the Codequiry CLI stored; it never spends a check. test/remote.js drives the signed-in server with the official SDK's Streamable HTTP client, given CODEQUIRY_MCP_URL and a bearer token in CODEQUIRY_MCP_TOKEN.

Links

• Setup guide and docs
• REST API reference
• Node SDK and CLI

MIT licensed.