@devpick/mcp-server

Developer tools for AI agents. Merge PDFs, calculate subnets, test regex, diff JSON — things your AI agent can't do alone.

43 tools • 100% local • No API keys • Works offline

From devpick.sh — developer tools that don't suck.

Install

Claude Code

claude mcp add devpick -- npx -y @devpick/mcp-server

Cursor

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "devpick": {
      "command": "npx",
      "args": ["-y", "@devpick/mcp-server"]
    }
  }
}

VS Code (Copilot)

Add to .vscode/mcp.json:

{
  "servers": {
    "devpick": {
      "command": "npx",
      "args": ["-y", "@devpick/mcp-server"]
    }
  }
}

Codex

codex mcp add devpick -- npx -y @devpick/mcp-server

Tools

📄 PDF Operations

Your AI agent can't manipulate PDF files. Now it can.

| Tool | Description | |------|-------------| | merge_pdfs | Merge multiple PDFs into one | | split_pdf | Extract specific pages from a PDF | | pdf_info | Get page count, metadata, dimensions | | rotate_pdf | Rotate pages by 90°, 180°, or 270° | | pdf_watermark | Add text watermark to every page | | remove_pdf_pages | Remove specific pages from a PDF |

🧮 Computation

Things LLMs get wrong — subnet math, regex execution, cron scheduling.

| Tool | Description | |------|-------------| | calculate_subnet | CIDR subnet calculator (network, broadcast, hosts) | | test_regex | Execute regex against test strings with match details | | parse_cron | Parse cron expression and show next N execution times | | parse_jwt | Decode JWT header + payload | | calculate_chmod | Unix permission octal ↔ symbolic conversion | | json_diff | Structural diff between two JSON objects | | text_diff | Line-by-line text comparison | | generate_hash | MD5, SHA-1, SHA-256, SHA-512 hashing |

✨ Format & Validate

| Tool | Description | |------|-------------| | format_json | Prettify or minify JSON | | format_sql | Beautify SQL queries | | format_xml | Format XML documents | | format_html | Beautify HTML | | format_css | Format CSS | | validate_yaml | Validate YAML syntax | | validate_toml | Validate TOML syntax |

🔐 Encode & Decode

| Tool | Description | |------|-------------| | encode_base64 / decode_base64 | Base64 encoding/decoding | | encode_url / decode_url | URL encoding/decoding | | encode_html_entities / decode_html_entities | HTML entity encoding | | escape_string / unescape_string | String escape sequences |

🔄 Convert

| Tool | Description | |------|-------------| | yaml_to_json / json_to_yaml | YAML ↔ JSON | | csv_to_json / json_to_csv | CSV ↔ JSON | | json_to_typescript | Generate TypeScript interfaces from JSON | | markdown_to_html | Convert Markdown to HTML | | number_base_convert | Convert between binary, octal, decimal, hex | | hex_to_rgb / rgb_to_hex | Color format conversion |

🔧 Generate

| Tool | Description | |------|-------------| | generate_uuid | Generate UUID v4 | | generate_lorem_ipsum | Generate placeholder text | | generate_slug | URL-friendly slug from text | | generate_cron_expression | Build cron expressions from description | | count_words | Word, character, sentence, paragraph count |

Why?

Your AI coding agent can format JSON. It can encode Base64. But it can't:

• Merge your PDF files
• Calculate exact subnet boundaries without guessing
• Execute a regex against 50 test strings and report matches
• Diff two large JSON objects structurally

This MCP server gives your agent real tools for real work. Everything runs locally on your machine — no API keys, no network calls, no rate limits.

License

MIT — devpick.sh