openclaw - 💡(How to fix) Fix Regression: embedded agent prep takes 50–80s on 2026.4.29 even with only one channel plugin enabled [1 comments, 2 participants]

2026-05-02T09:26:33.820Z warn agent/embedded {"subsystem":"agent/embedded"} [trace:embedded-run] startup stages: runId=06cf5a7d-cc14-4af0-a10d-c8d1d1ecf0c5 sessionId=c16878ff-5573-444b-bf31-c359ad785901 phase=attempt-dispatch totalMs=20525 stages=workspace:0ms@0ms,runtime-plugins:1ms@1ms,hooks:1ms@2ms,model-resolution:3545ms@3547ms,auth:7600ms@11147ms,context-engine:0ms@11147ms,attempt-dispatch:9378ms@20525ms 2026-05-02T09:27:51.771Z warn agent/embedded {"subsystem":"agent/embedded"} [trace:embedded-run] prep stages: runId=06cf5a7d-cc14-4af0-a10d-c8d1d1ecf0c5 sessionId=c16878ff-5573-444b-bf31-c359ad785901 phase=stream-ready totalMs=77946 stages=workspace-sandbox:9463ms@9463ms,skills:0ms@9463ms,core-plugin-tools:9803ms@19266ms,bootstrap-context:8611ms@27877ms,bundle-tools:5799ms@33676ms,system-prompt:25960ms@59636ms,session-resource-loader:1909ms@61545ms,agent-session:2ms@61547ms,stream-setup:16399ms@77946ms 2026-05-02T09:28:03.548Z warn diagnostic {"subsystem":"diagnostic"} liveness warning: reasons=event_loop_delay,event_loop_utilization,cpu interval=36s eventLoopDelayP99Ms=28336.7 eventLoopDelayMaxMs=28336.7 eventLoopUtilization=1 cpuCoreRatio=1.01 active=1 waiting=0 queued=1 2026-05-02T09:30:43.011Z warn agent/embedded {"subsystem":"agent/embedded"} [trace:embedded-run] startup stages: runId=c7565319-d7f5-4de6-bba1-d31b39cd96ef sessionId=c16878ff-5573-444b-bf31-c359ad785901 phase=attempt-dispatch totalMs=20210 stages=workspace:0ms@0ms,runtime-plugins:2ms@2ms,hooks:0ms@2ms,model-resolution:3546ms@3548ms,auth:7625ms@11173ms,context-engine:1ms@11174ms,attempt-dispatch:9036ms@20210ms 2026-05-02T09:31:34.824Z warn agent/embedded {"subsystem":"agent/embedded"} [trace:embedded-run] prep stages: runId=c7565319-d7f5-4de6-bba1-d31b39cd96ef sessionId=c16878ff-5573-444b-bf31-c359ad785901 phase=stream-ready totalMs=51809 stages=workspace-sandbox:387ms@387ms,skills:0ms@387ms,core-plugin-tools:9807ms@10194ms,bootstrap-context:28ms@10222ms,bundle-tools:5785ms@16007ms,system-prompt:17573ms@33580ms,session-resource-loader:1873ms@35453ms,agent-session:2ms@35455ms,stream-setup:16353ms@51808ms 2026-05-02T09:34:59.076Z warn agent/embedded {"subsystem":"agent/embedded"} [trace:embedded-run] startup stages: runId=ef85fcec-6bb3-45c6-89e0-33bf2ac87bbf sessionId=c16878ff-5573-444b-bf31-c359ad785901 phase=attempt-dispatch totalMs=23588 stages=workspace:0ms@0ms,runtime-plugins:2ms@2ms,hooks:0ms@2ms,model-resolution:3590ms@3592ms,auth:7737ms@11329ms,context-engine:0ms@11329ms,attempt-dispatch:12259ms@23588ms 2026-05-02T09:35:51.284Z warn agent/embedded {"subsystem":"agent/embedded"} [trace:embedded-run] prep stages: runId=ef85fcec-6bb3-45c6-89e0-33bf2ac87bbf sessionId=c16878ff-5573-444b-bf31-c359ad785901 phase=stream-ready totalMs=52204 stages=workspace-sandbox:5ms@5ms,skills:1ms@6ms,core-plugin-tools:9913ms@9919ms,bootstrap-context:37ms@9956ms,bundle-tools:5885ms@15841ms,system-prompt:17822ms@33663ms,session-resource-loader:1901ms@35564ms,agent-session:3ms@35567ms,stream-setup:16636ms@52203ms

TL;DR

The embedded agent startup/prep delay can be mitigated by investigating and optimizing the expensive stages in the agent prep process, such as core-plugin-tools, bundle-tools, system-prompt, and stream-setup.

Guidance

• Investigate the recent changes in 2026.4.29 that may have introduced a regression in the embedded agent prep process, focusing on the expensive stages identified.
• Consider implementing caching or prebuilding for the expensive stages, such as core-plugin-tools, bundle-tools, system-prompt, and stream-setup, to reduce the delay.
• Review the event-loop diagnostics to understand the cause of the severe blocking and high eventLoopUtilization.
• Test the embedded agent with different configurations and workloads to identify potential bottlenecks and areas for optimization.

Example

No specific code snippet can be provided without more information on the implementation details of the embedded agent and the expensive stages. However, a potential optimization could involve caching the results of the core-plugin-tools stage to avoid redundant computations.

Notes

The issue seems to be related to the recent update to 2026.4.29, and the expensive stages identified are consistent across different tests. However, without more information on the implementation details, it is difficult to provide a more specific solution.

Recommendation

Apply a workaround by investigating and optimizing the expensive stages in the agent prep process, and consider implementing caching or prebuilding to reduce the delay. This approach is recommended because it addresses the root cause of the issue and can provide a significant improvement in performance.