How much of what your coding agent spends is actually work: the share of paid tokens that became output, not context re-reading. Local, no network calls.
Setup
Choose your client, copy the config, then fill in any required environment variables.
claude mcp add contextburn -- uvx contextburn~/.codex/config.toml
[mcp_servers.contextburn]
command = "uvx"
args = ["contextburn"]
# codex mcp add contextburn -- uvx contextburn~/.cursor/mcp.json
{
"mcpServers": {
"contextburn": {
"command": "uvx",
"args": [
"contextburn"
]
}
}
}.vscode/mcp.json
{
"servers": {
"contextburn": {
"type": "stdio",
"command": "uvx",
"args": [
"contextburn"
]
}
}
}~/.gemini/settings.json
{
"mcpServers": {
"contextburn": {
"command": "uvx",
"args": [
"contextburn"
]
}
}
}claude_desktop_config.json
{
"mcpServers": {
"contextburn": {
"command": "uvx",
"args": [
"contextburn"
]
}
}
}.trae/mcp.json
{
"mcpServers": {
"contextburn": {
"command": "uvx",
"args": [
"contextburn"
]
}
}
}