Most coding agents grep and glob raw source into LLM's context window and hope the model is smart. Benzi compiles your codebase into a queryable map of calls, data flow, control flow and class hierarchy first; then it reads, explains, runs, edits and verifies your code using it.
Setup
Choose your client, copy the config, then fill in any required environment variables.
claude mcp add benzi -- uvx benzi-mcp~/.codex/config.toml
[mcp_servers.benzi]
command = "uvx"
args = ["benzi-mcp"]
# codex mcp add benzi -- uvx benzi-mcp~/.cursor/mcp.json
{
"mcpServers": {
"benzi": {
"command": "uvx",
"args": [
"benzi-mcp"
]
}
}
}.vscode/mcp.json
{
"servers": {
"benzi": {
"type": "stdio",
"command": "uvx",
"args": [
"benzi-mcp"
]
}
}
}~/.gemini/settings.json
{
"mcpServers": {
"benzi": {
"command": "uvx",
"args": [
"benzi-mcp"
]
}
}
}claude_desktop_config.json
{
"mcpServers": {
"benzi": {
"command": "uvx",
"args": [
"benzi-mcp"
]
}
}
}.trae/mcp.json
{
"mcpServers": {
"benzi": {
"command": "uvx",
"args": [
"benzi-mcp"
]
}
}
}