openclaw - 💡(How to fix) Fix Per-session core-plugin-tools init (~32s) exceeds full boot (~15s) — plugin tools re-initialized per session

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…

Every new agent session pays a ~25–45s core-plugin-tools prep stage. This is longer than booting the entire gateway, which loads all plugins in ~15s. Plugin tools appear to be fully re-initialized on each session prep rather than reused from the warm boot state.

Root Cause

Every new agent session pays a ~25–45s core-plugin-tools prep stage. This is longer than booting the entire gateway, which loads all plugins in ~15s. Plugin tools appear to be fully re-initialized on each session prep rather than reused from the warm boot state.

Fix Action

Fix / Workaround

Workaround today

None at the config layer (verified). Session reuse keeps follow-ups in the same session warm; only first-turns pay the full init.

Code Example

http server listening (4 plugins: browser, discord, memory-core, memory-wiki; 15.1s)

---

prep stages ... core-plugin-tools:31966ms ... system-prompt:9377ms ... stream-setup:8897ms ... totalMs=56633
RAW_BUFFERClick to expand / collapse

openclaw version: 2026.4.29

Summary

Every new agent session pays a ~25–45s core-plugin-tools prep stage. This is longer than booting the entire gateway, which loads all plugins in ~15s. Plugin tools appear to be fully re-initialized on each session prep rather than reused from the warm boot state.

Evidence (embedded-run stage traces, main agent)

Boot loads all plugins quickly:

http server listening (4 plugins: browser, discord, memory-core, memory-wiki; 15.1s)

But every per-session prep re-pays more than that:

prep stages ... core-plugin-tools:31966ms ... system-prompt:9377ms ... stream-setup:8897ms ... totalMs=56633

core-plugin-tools is the dominant cost of every conversational turn.

What I ruled out (not plugin-specific, not config-fixable)

  • Removing browser from plugins.allow → still ~31.8s (and breaks the openclaw browser CLI).
  • Setting memory-wiki search.backend: local → still ~31.9s, with memory-core/lancedb still staging.
  • The repeated [plugins] <name> staging bundled runtime deps (33 specs) / installed ... in 6ms lines fire per session — install is cached/fast; the cost is init/activation, not dependency install.
  • plugins.entries.<name>.activation is rejected by the config schema, so there's no user-facing lazy/eager toggle for bundled plugins.

Impact

Conversational turns average 30–65s (max ~150s). With per-message threading (auto-thread), every fresh topic is a new (cold) session and pays the full init. Under concurrency, simultaneous agent runs can exceed the gateway timeout and fall back to embedded.

Asks

  1. Cache/reuse plugin-tool initialization across sessions (warm-pool) instead of re-initializing per session.
  2. Or expose a lazy/eager activation control for bundled plugins (e.g. plugins.entries.<name>.activation).

Workaround today

None at the config layer (verified). Session reuse keeps follow-ups in the same session warm; only first-turns pay the full init.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING