Tools
Tools let the model act on the current workspace and connected services. The active runtime, permission policy, selected model, and available integrations determine which tools can execute.
Built-in tools
| Area | Tools |
|---|---|
| Find and read files | glob_files, grep_files, read_file |
| Write and edit files | write_file, edit_file |
| Commands | shell |
| Web | web_search, web_fetch |
| Images | vision |
| Skills | skill, install_skill |
| Subagents | subagent |
| MCP | capability_search, mcp_select_tool, mcp_features, plus selected server tools |
| Interaction and runtime | ask_user_question, read_tool_result |
Use glob_files to find paths and grep_files to find literal text. capability_search finds installed skills and configured MCP tools. Matching MCP schemas load automatically within the schema budget; mcp_select_tool can select one explicitly.
nod does not currently include interactive browser or CDP tools.
See Web search for search selection, source scoping, citations, permissions, privacy, and the difference between web_search and web_fetch.
Background commands
The shell tool runs commands, including development servers and file watchers. When a command is still running, nod keeps its execution handle so the agent can check output, send input, or stop it without starting it again.
The tool has three actions: run starts a command, interact checks output or sends input, and stop ends it. Ask nod to start a service, check its progress, or stop it in the conversation. /clear keeps running commands for the workspace; /reset stops and forgets them.
Background commands are operating system processes. Subagents are separate nod sessions for delegated reasoning and tool work.
Large tool results
Large results are kept out of the immediate model response so one command or search does not consume the entire context window.
| Stage | What nod provides |
|---|---|
| Initial result | A bounded preview, the retained byte count, and a session-scoped handle. |
| Follow-up inspection | read_tool_result can read a byte range or search the retained result for a literal query. |
| Saved session | The redacted result remains outside session.json; its handle stays attached to the tool step. |
The model calls read_tool_result when it needs evidence beyond the preview. Handles must be copied exactly and work only with their session's tool-result store. The max_tool_result_bytes configuration field controls how many bytes nod retains from one tool result.
Vision
The vision tool inspects images on behalf of the model in a separate request. See Vision for how images are routed.
Scope and permissions
File and command tools start from the primary workspace. Active additional directories extend that scope, but they do not bypass permission rules.
ACP ask and code modes can use the full runtime set. Tool availability can still vary by entrypoint, model capability, platform, and integration state.
Tool calls are policy checked
A model choosing a tool does not guarantee execution. nod evaluates effective permission rules, session grants, mode, and workspace access first.