Skip to content

llmist mcp import-claude-code

Reads ~/.claude.json (or $CLAUDE_CONFIG_HOME) and emits TOML [mcp.servers.<name>] blocks you can paste — or --write directly — into your llmist config.

Why

Most users running llmist already have MCP servers configured in Claude Code, Cursor, or Cline (which all share the ~/.claude.json shape). Retyping that config is friction; this command copies it.

Usage

llmist mcp import-claude-code              # print blocks to stdout
llmist mcp import-claude-code >> ~/.llmist/config.toml
llmist mcp import-claude-code --write      # append directly to ~/.llmist/config.toml
llmist mcp import-claude-code --write /path/to/llmist.toml
llmist mcp import-claude-code --source /custom/claude.json

What gets imported

The command recognizes both top-level mcpServers and per-project projects[*].mcpServers shapes that Claude Code uses.

For each entry it emits the matching llmist TOML block:

Claude Code shape	llmist block
`{ command, args?, env? }`	`[mcp.servers.<n>] transport = "stdio" command = ... args = ... env = ...`
`{ type: "http" \| "streamable-http" \| "sse", url, headers? }`	`[mcp.servers.<n>] transport = "http" url = ... headers = ...`
anything else	skipped with a stderr warning

What does NOT get imported

Claude-Code-specific keys outside the MCP slot (chat profiles, hooks).
Auth tokens stored in OS keychains rather than the JSON file.
Server entries with enabled = false — these aren’t represented in Claude Code’s shape; you can disable per-block via enabled = false after import.

Review before keeping

The output is prefixed with a banner reminding you to review. The recommended workflow:

llmist mcp import-claude-code > /tmp/imported.toml
# Inspect, prune, edit
cat /tmp/imported.toml >> ~/.llmist/config.toml