Logging

In plain terms

logLevel is how chatty the agent is in your terminal.

"silent" (default) — hush
"info" — “run started / ended”
"debug" — also print tool calls

const agent = new Agent({
  name: "assistant",
  model: openai(process.env.OPENAI_MODEL ?? "gpt-6.1-sol"),
  logLevel: "info",
});

Logger Class

The Logger is used internally by agents. You can also instantiate it directly:

import { Logger } from "@agentium/core";

const logger = new Logger({
  level: "info",
  color: true,
  prefix: "myapp",
});

logger.info("Agent started");
logger.debug("Tool call", { toolName: "getWeather", args: { city: "Tokyo" } });

LogLevel

Level	Order	Use Case
`debug`	0	Tool call details, arguments, results
`info`	1	Agent start/end, token usage, duration
`warn`	2	Warnings
`error`	3	Errors
`silent`	4	No logging (default for agents)

Setting logLevel in AgentConfig

Control agent logging with the logLevel property:

const agent = new Agent({
  name: "ToolBot",
  model: openai(process.env.OPENAI_MODEL ?? "gpt-6.1-sol"),
  tools: [weatherTool],
  logLevel: "debug",  // "debug" | "info" | "warn" | "error" | "silent"
});

"debug" — Full tool names, arguments, and results (JSON pretty-printed, not sliced at 200 chars).
"info" — Agent start, output, token usage, and duration.
"silent" — No logs (default).

What Gets Logged

Agent start

When a run begins: agent name and input preview.

Tool calls

At debug level: tool name and arguments.

Tool results

At debug level: full tool result. JSON strings are pretty-printed. Bodies over 100k characters end with … (N more chars) instead of a silent mid-JSON cut.

Token usage

At info level: prompt tokens, completion tokens, total tokens.

Duration

At info level: wall-clock duration (ms or seconds).

Agent end

When a run completes: output preview and summary.

LoggerConfig

Property	Type	Default	Description
`level`	`LogLevel`	`"info"`	Minimum level to log.
`color`	`boolean`	`process.stdout.isTTY !== false`	Enable colored output. Auto-disabled when not a TTY.
`prefix`	`string`	`"agentium"`	Prefix shown in log lines. Agents use their `name` as prefix.
`maxPayloadChars`	`number`	`100000`	Cap for a single tool arg/result body at debug. Raise if you need megabyte dumps; lower if a toolkit is noisy.

Colored Terminal Output

The logger uses ANSI escape codes for colored output:

Cyan — Agent name, box borders
Magenta — Tool names
Green — Token counts, success indicators
Yellow — Duration, string values
Gray — Timestamps, metadata
Red — Errors

Box-drawing characters (┌, │, └, ─) create a structured layout for agent runs.

Example Output

With logLevel: "info":

┌─ Agent: ToolBot ───────────────────────────────────────────────────────
│ Input:   What's the weather in Tokyo? Also, what's 42 * 17?
│
│ Output:  Tokyo is 22°C and sunny. 42 * 17 = 714.
│
│ Tokens: ↑ 245  ↓ 38  Σ 283
│ Duration: 1.2s
└────────────────────────────────────────────────────────────────────────

With logLevel: "debug", each tool is two compact lines — args stay on one line when small, pretty-printed when not:

│ → getWeather  {"city":"Tokyo"}
│ ← getWeather  Tokyo: 22°C, sunny
│ → search_orders
│   {
│     "query": "delayed",
│     "limit": 50
│   }
│ ← search_orders
│   {
│     "orders": [ ... ],
│     "total": 12
│   }

Example: Enabling Logging

const agent = new Agent({
  name: "assistant",
  model: openai(process.env.OPENAI_MODEL ?? "gpt-6.1-sol"),
  logLevel: "info",
});

const agent = new Agent({
  name: "ToolBot",
  model: openai(process.env.OPENAI_MODEL ?? "gpt-6.1-sol"),
  tools: [weatherTool, calculator],
  logLevel: "debug",
});

const agent = new Agent({
  name: "assistant",
  model: openai(process.env.OPENAI_MODEL ?? "gpt-6.1-sol"),
  logLevel: "silent",  // or omit — default is "silent"
});

Build

Agent applications

Harness runtime

Teams and workflows

Data and capabilities

Interfaces and jobs

Voice and telephony

Verify and manage

In plain terms

Logger Class

LogLevel

Setting logLevel in AgentConfig

What Gets Logged

LoggerConfig

Colored Terminal Output

Example Output

Example: Enabling Logging

​In plain terms

​Logger Class

​LogLevel

​Setting logLevel in AgentConfig

​What Gets Logged

​LoggerConfig

​Colored Terminal Output

​Example Output

​Example: Enabling Logging

In plain terms

Logger Class

LogLevel

Setting logLevel in AgentConfig

What Gets Logged

LoggerConfig

Colored Terminal Output

Example Output

Example: Enabling Logging