Reading Your Agent Logs

Your agent logs are the single most useful tool for understanding what's happening inside your agent. This guide shows you how to find them, read them, and use them to solve problems.

Access Your Logs

1. Open the Agento dashboard and click on your agent
2. Click the Logs tab

The logs show a chronological feed of events from your agent's container. New entries appear at the top in real time while the agent is running.

Understand Log Levels

Each log entry has a level that tells you its severity:

Level	Meaning	Action
INFO	Normal operation — startup messages, session events, model calls	No action needed
WARN	Something unexpected but non-fatal — slow responses, retries	Monitor; may indicate a developing issue
ERROR	Something failed — credential errors, API failures, crashes	Investigate and fix

Most of the time you'll see INFO-level entries. Focus on WARN and ERROR entries when debugging.

Common Log Patterns

Successful Startup

INFO  Agent starting...
INFO  Loading configuration
INFO  Provider: anthropic (claude-sonnet-4-5-20250929)
INFO  Channels: telegram
INFO  Agent ready

This means everything initialized correctly. The agent is running and accepting messages.

Credential Error

ERROR Provider authentication failed: Invalid API key
ERROR Agent failed to start

The API key is wrong or expired. Go to Settings and re-enter your credentials.

Rate Limiting

WARN  Rate limit hit, retrying in 5s...
WARN  Rate limit hit, retrying in 15s...
ERROR Max retries exceeded for API call

Your AI provider is throttling requests. This happens during heavy usage. The agent retries automatically, but if it exhausts retries, the message will fail. Consider switching to a faster model or upgrading your provider plan.

Session Activity

INFO  New session: sess_abc123
INFO  Message received (telegram, user: 12345678)
INFO  Model response: 247 tokens, 1.3s

Normal message flow. You can see where messages come from (telegram, web chat), token counts, and response latency.

Container Errors

ERROR Command failed: sudo apt install -y nonexistent-pkg
ERROR E: Unable to locate package nonexistent-pkg

The agent tried to run a command that failed. This is usually harmless — the agent should recover and try a different approach.

Filter and Search Logs

Use the search bar at the top of the Logs tab to filter entries. Useful searches:

• ERROR — show only errors
• rate limit — find throttling events
• session — track session lifecycle
• token — see token usage per message

Copy Logs for Debugging

When you need help debugging, copying your logs gives the most context. Here's how:

Copy from the Dashboard

1. Open the Logs tab
2. Select the time range around the problem
3. Click Copy Logs (or select the text and Ctrl+C / Cmd+C)

Paste into an LLM for Analysis

Agent logs are structured enough that an AI model can analyze them effectively. Here's a prompt template you can use:

I'm running an AI agent on Agento (managed OpenClaw hosting).
The agent uses [Anthropic Claude / OpenAI GPT / Google Gemini] as its AI provider.

Problem: [describe what's happening]

Here are the agent logs from around the time of the issue:

[paste your logs here]

Questions:
1. What is causing this error?
2. How can I fix it?
3. Is this a configuration issue or a provider issue?

This works well with any AI model — Claude, ChatGPT, or Gemini. The structured log format makes it easy for models to identify patterns and root causes.

What to Include

When sharing logs for debugging:

• Include 2-3 minutes of context before the error, not just the error line itself
• Include the startup sequence if the agent failed to start
• Redact sensitive data if sharing publicly — API keys should never appear in logs, but usernames or message content might

Monitor Performance

Logs can help you optimize your agent's performance:

Track response latency. Look for Model response: X tokens, Y.Zs entries. If latency is consistently high:

• Switch to a faster model
• Shorten your system prompt
• Reduce conversation history length

Track token usage. High token counts per response increase cost and latency. If your agent is generating very long responses, add instructions like "Keep responses under 200 words" to your system prompt.

Watch for retries. Frequent WARN entries about retries suggest you're close to your provider's rate limits. Consider upgrading your API tier.

Log Retention

Agento retains logs for the lifetime of the agent's current container. Logs from previous container instances are not preserved. If you need long-term log storage, copy important logs before restarting or upgrading your agent.

Troubleshooting Log Access

Logs tab is empty: The agent may not have started yet, or it may have crashed immediately. Check the agent's status on the dashboard.

Logs stop updating: Try refreshing the page. If logs still don't update, the agent may have stopped — check its status.

Can't find the error: Use the search bar to filter for ERROR or WARN. If there are no errors in the logs but something is wrong, the issue may be in the system prompt or model behavior rather than the infrastructure.