Configuration
kerf.toml
Optional project-level config file created by kerf init.
[server]
| Key | Type | Default | Description |
|---|---|---|---|
host |
string | "0.0.0.0" |
Server bind address |
port |
int | 8000 |
Server port |
[defaults]
| Key | Type | Default | Description |
|---|---|---|---|
fallback |
string | "retry" |
Default fallback policy for workflows that don't specify one |
If kerf.toml is missing, all defaults apply. If it has a parse error, Kerf warns and falls back to defaults.
Python 3.10
Python 3.10 needs the tomli package to parse TOML (included automatically as a dependency). Python 3.11+ uses the built-in tomllib.
mcp.json
Optional MCP server configuration. When present, Kerf passes --mcp-config mcp.json to Claude CLI calls, giving the LLM access to external tools during reasoning.
{
"mcpServers": {
"postgres": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-postgres", "postgresql://localhost/mydb"],
"env": {}
}
}
}
Manage entries with kerf add mcp <name>. See Using MCP Servers for details.
Project detection
Kerf finds your project root by walking up from the current directory looking for a .kerf marker file. This works the same way git finds .git/, so you can run kerf commands from any subdirectory.
If no .kerf file is found, Kerf uses the current directory as the project root.