Nils CodexInstall

Summary

Runs an OpenAI Codex agent for each incoming item using the required Prompt, with optional system instructions and installed or inline skills. Configure its model, reasoning, thread behavior, filesystem sandbox, and workspace; authentication must be set up beforehand with the Nils Codex Setup node.

Inputs

• Prompt (required) — Instruction sent to Codex. By default, reads chatInput, prompt, or text from the incoming item.
• System Prompt — Static instructions prepended to each run.
• Static Skills — Comma- or newline-separated names of installed skills to apply.
• Dynamic Skills — Per-run skill names, inline Markdown, or skill objects supplied as an array or object.
• Model — Selects GPT-5.5, GPT-5.4, or GPT-5.4 Mini.
• Reasoning Effort — Sets reasoning effort to extra high, high, low, or medium.
• Personality — Sets the Codex personality to default, friendly, or pragmatic.
• Thread Mode — Starts a new thread, continues the thread ID saved in node static data, or resumes by ID. For resume mode, provide Thread ID.
• Thread ID — Thread to resume when Thread Mode is Resume Thread By ID.
• Sandbox — Sets Codex filesystem access to read only, workspace write, or full access.
• Working Directory — Optional directory Codex should use as its workspace.
• Options — Optional JSON output schema, progress logging to the execution log, and timeout in seconds (default 300; minimum 30).

Output shape

One output item per input item, each containing a JSON object.

Successful items include output, threadId, items, usage, model, finishReason, and appliedSkills. With Continue On Fail enabled, failed items contain an error field. Streaming progress logs completed agent messages; its returned items and usage are empty/null. Continuing a previous thread stores its threadId in node static data between runs.

Links

• n8n community nodes installation guide