nexus_agent_run
Run an autonomous agent that reads files, writes code, executes commands, and searches the web, iterating until the task is done, returning the final answer and full action trace.
Instructions
Run an autonomous AI agent that can read files, write code, execute shell commands, search the web, and iterate until the task is complete. The agent uses the specified model (GPT-5.4, Nemotron, Gemini, etc.) and works independently through a ReAct loop. Returns the final answer plus a full trace of every action taken.
Input Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| role | No | Optional label for this agent (e.g., "coder", "reviewer") | |
| task | Yes | What the agent should accomplish. Be specific. | |
| model | Yes | Model ID to use (e.g., "gpt-5.4", "nvidia/nemotron-3-super-120b-a12b") | |
| provider | Yes | Provider ID (e.g., "openai", "nvidia", "gemini", "groq") | |
| maxTimeMs | No | Total timeout in milliseconds (default: 300000 = 5 minutes) | |
| maxTokens | No | Max tokens per model call | |
| workingDir | No | Directory the agent can read/write files in (default: current project directory) | |
| temperature | No | Sampling temperature (0.0-2.0) | |
| enabledTools | No | Tools to enable. Available: read_file, write_file, edit_file, list_directory, search_files, run_command, web_fetch, think. Default: all tools. | |
| systemPrompt | No | Custom agent persona or additional instructions | |
| maxIterations | No | Max agent loop iterations (default: 25) |