@yadimon/claude-code-to-llm

Minimal SDK and CLI wrapper around Claude Code headless mode for raw prompt requests.

Install

npm install @yadimon/claude-code-to-llm

Requirements:

• Node.js >=20
• installed claude CLI in PATH or CLAUDE_CODE_TO_LLM_CLI_PATH
• valid Claude Code login on the machine

The wrapper defaults to the Claude Code auth bundle at:

• ~/.claude.json
• ~/.claude/.credentials.json

If you need to verify that Claude Code is currently logged in, run:

claude auth status

What It Provides

• a small SDK for raw prompt execution with minimal prompt overhead
• a CLI for direct prompt mode from flags, files, or stdin
• structured streaming events for adapters such as HTTP compatibility servers
• isolated execution via a temporary home directory copied from your Claude Code auth bundle

SDK

import { runPrompt } from "@yadimon/claude-code-to-llm";

const result = await runPrompt("Hello", {
  model: "claude-sonnet-4-6",
  maxTokens: 256
});

console.log(result.content);
console.log(result.usage);

For streamed events:

import { streamPrompt } from "@yadimon/claude-code-to-llm";

for await (const event of streamPrompt("Hello", {
  model: "claude-sonnet-4-6"
})) {
  if (event.type === "response.output_text.delta") {
    process.stdout.write(event.delta);
  }
}

CLI

claude-code-to-llm --prompt "Hello"
claude-code-to-llm --input-file ./prompt.txt --json
cat ./prompt.txt | claude-code-to-llm --stream --json

Supported CLI options:

--prompt <text>
--input-file <path>
--stream
--json
--model <name>
--reasoning-effort <level>
--max-tokens <n>
--search
--auth-path <path>
--credentials-path <path>
--settings-path <path>
--config-home <path>
--cwd <path>
--cli <path>

Runtime Configuration

| Variable | Default | Description | |---|---|---| | CLAUDE_CODE_TO_LLM_AUTH_PATH | ~/.claude.json | Path to the Claude Code session file. | | CLAUDE_CODE_TO_LLM_CREDENTIALS_PATH | ~/.claude/.credentials.json | Path to the Claude Code credentials file. | | CLAUDE_CODE_TO_LLM_SETTINGS_PATH | ~/.claude/settings.json when present | Optional Claude Code settings file to copy into the temp home. | | CLAUDE_CODE_TO_LLM_CLI_PATH | claude | Path to the Claude Code CLI binary. | | CLAUDE_CODE_TO_LLM_REASONING_EFFORT | low | Default reasoning effort passed to Claude Code. | | CLAUDE_CODE_TO_LLM_CONFIG_HOME | temp dir | Temporary Claude Code home directory for a run. | | CLAUDE_CODE_TO_LLM_WORKSPACE | temp dir | Workspace directory used for execution. | | CLAUDE_CODE_TO_LLM_LOCAL_HOME | .claude-code-to-llm/ | Local directory used by the auth copy helper. |

Notes

• The wrapper calls claude --print --output-format stream-json.
• maxTokens maps to CLAUDE_CODE_MAX_OUTPUT_TOKENS for the spawned Claude Code process.
• The package is intentionally focused on raw prompt execution. It does not expose Claude Code tools through its public API.
• Web search is off by default. The wrapper always forces the WebSearch permission via --allowed-tools WebSearch (when webSearch: true / --search) or --disallowed-tools WebSearch, overriding any WebSearch entry in settings.json.

Development

npm run build --workspace @yadimon/claude-code-to-llm
npm run lint --workspace @yadimon/claude-code-to-llm
npm run typecheck --workspace @yadimon/claude-code-to-llm