Problem Description
The agenticworkflows logs MCP tool times out with context deadline exceeded when filtering by engine (e.g. engine: "claude") even with a small count. The default timeout appears to be 60s, which is insufficient for engine-filtered queries — the underlying implementation likely has to scan/paginate through many workflow runs to find ones matching the engine rather than filtering more efficiently.
Command/Tool
- Tool: logs (agentic-workflows MCP server)
- Command:
printf '{"engine":"claude","count":2,"max_tokens":3000}' | agenticworkflows logs .
Steps to Reproduce
- Call the
logs tool with engine: "claude" and count: 2 (default timeout, no explicit timeout param).
- Observe the call runs for the full 60s and fails.
- Retry the same query with
timeout: 300 explicitly set — it succeeds in ~30s.
Expected Behavior
Engine-filtered log queries with a small count (e.g. 2) should complete well within the default timeout, similar to unfiltered or workflow_name-filtered queries which typically complete in 6-11s.
Actual Behavior
With default timeout, the call errors out after 60s with:
Error [0]: context deadline exceeded
Retrying with timeout: 300 succeeds after ~30s, confirming the query itself works but the default timeout budget is too aggressive for engine-filtered queries.
Environment
- Repository: github/gh-aw
- Run ID: 31461891659
- Date: 2026-08-11
Impact
- Severity: Medium
- Frequency: Likely always for engine-filtered queries against a repo with many workflow runs (this repo has 284 workflows across 12 engines)
- Workaround: Explicitly pass a higher
timeout (e.g. timeout: 300) when using the engine filter.
Additional Context
Other logs filters tested (workflow_name, start_date incl. future dates, nonexistent workflow name) all completed quickly (<11s) and returned clear, well-formed results/error messages. Only the engine filter exhibited this default-timeout issue. Consider raising the default timeout for engine-filtered queries, optimizing the underlying scan (e.g. using GitHub Actions workflow-run list filters server-side), or adding partial/streaming results before the deadline.
A minor, non-blocking usability gap was also found: the compile tool's --help text does not clarify that workflows must be an array (passing a plain string yields a JSON-schema validation error: type: ... has type "string", want one of "null, array"). Adding an example with array syntax to the help text would prevent this trial-and-error.
Generated by 🧪 Daily Cli Tools Tester · auto · 48.2 AIC · ⌖ 5.2 AIC · ⊞ 13.5K · ◷
Problem Description
The
agenticworkflows logsMCP tool times out withcontext deadline exceededwhen filtering byengine(e.g.engine: "claude") even with a smallcount. The default timeout appears to be 60s, which is insufficient for engine-filtered queries — the underlying implementation likely has to scan/paginate through many workflow runs to find ones matching the engine rather than filtering more efficiently.Command/Tool
Steps to Reproduce
logstool withengine: "claude"andcount: 2(default timeout, no explicittimeoutparam).timeout: 300explicitly set — it succeeds in ~30s.Expected Behavior
Engine-filtered log queries with a small
count(e.g. 2) should complete well within the default timeout, similar to unfiltered or workflow_name-filtered queries which typically complete in 6-11s.Actual Behavior
With default timeout, the call errors out after 60s with:
Retrying with
timeout: 300succeeds after ~30s, confirming the query itself works but the default timeout budget is too aggressive for engine-filtered queries.Environment
Impact
timeout(e.g.timeout: 300) when using theenginefilter.Additional Context
Other logs filters tested (workflow_name, start_date incl. future dates, nonexistent workflow name) all completed quickly (<11s) and returned clear, well-formed results/error messages. Only the
enginefilter exhibited this default-timeout issue. Consider raising the default timeout for engine-filtered queries, optimizing the underlying scan (e.g. using GitHub Actions workflow-run list filters server-side), or adding partial/streaming results before the deadline.A minor, non-blocking usability gap was also found: the
compiletool's--helptext does not clarify thatworkflowsmust be an array (passing a plain string yields a JSON-schema validation error:type: ... has type "string", want one of "null, array"). Adding an example with array syntax to the help text would prevent this trial-and-error.