observe_logs_search

Tool

Search the log lines of the current product, newest first.

Search the log lines of the current product, newest first. Filters are structured; there is no SQL surface. Page with nextCursor.

Name	Type	Required	Description
q	string	—	Lucene-style expression, ANDed with the structured filters. `key:value` (service, environment, severity, trace_id, span_id, body; any other key matches an attribute), "quoted phrases", AND / OR / NOT in uppercase, parentheses. A bare word matches a whole word in the message body. Example: service:api AND NOT severity:INFO AND "connection refused"
to	string	—	Newest timestamp to include, ISO 8601
from	string	—	Oldest timestamp to include, ISO 8601
limit	integer	—	How many lines to return, 1-1000, default 100
cursor	string	—	The nextCursor of the previous page; omit for the first page
search	string	—	Match the message body. A single word matches whole words and is indexed; anything longer is a case-insensitive substring match
service	string	—	Only lines of this service, e.g. "api"
traceId	string	—	Only lines of this trace
severityIn	array	—	Only lines whose severity is one of these, e.g. ["ERROR", "FATAL"]
environment	string	—	Only lines of this environment, e.g. "production"

Field	Type	Always	Description
rows	array	•	The matching log lines, newest first
nextCursor	string	•	Pass back as cursor for the next page; null when the last page was reached

Search the log lines of the current product, newest first.

When to use

Search the log lines of the current product, newest first. Filters are structured; there is no SQL surface. Page with nextCursor.

Ask your agent something like this:

Search the log lines of the current product, newest first.

The agent will invoke observe_logs_search with these arguments:

{}