CLI, MCP server, and TypeScript library that extracts text, layout, and page images from PDFs for AI agents — when extraction goes wrong, it says so per page and says what to do next.
Setup
Choose your client, copy the config, then fill in any required environment variables.
claude mcp add pdfvision -- npx -y pdfvision mcp~/.codex/config.toml
[mcp_servers.pdfvision]
command = "npx"
args = ["-y", "pdfvision", "mcp"]
# codex mcp add pdfvision -- npx -y pdfvision mcp~/.cursor/mcp.json
{
"mcpServers": {
"pdfvision": {
"command": "npx",
"args": [
"-y",
"pdfvision",
"mcp"
]
}
}
}.vscode/mcp.json
{
"servers": {
"pdfvision": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"pdfvision",
"mcp"
]
}
}
}~/.gemini/settings.json
{
"mcpServers": {
"pdfvision": {
"command": "npx",
"args": [
"-y",
"pdfvision",
"mcp"
]
}
}
}claude_desktop_config.json
{
"mcpServers": {
"pdfvision": {
"command": "npx",
"args": [
"-y",
"pdfvision",
"mcp"
]
}
}
}.trae/mcp.json
{
"mcpServers": {
"pdfvision": {
"command": "npx",
"args": [
"-y",
"pdfvision",
"mcp"
]
}
}
}