#94162: Performance: bundle-tools loading adds 6-7s latency on every agent request by mmyzwl · Pull Request #94230 · openclaw/openclaw

mmyzwl · 2026-06-17T16:33:33Z

Summary

Every agent request incurs ~6-7 seconds of latency before the model starts generating, with the bundle-tools prep stage accounting for 96-99% of total prep time. The bottleneck is that getCatalog() in SessionMcpRuntime connects to configured MCP servers sequentially — each server's connectWithTimeout + listAllToolsBestEffort runs one after another, so total latency is the sum of all servers' connection times.

With 4-5 MCP servers at the default 1,500ms tools/list timeout, that's 6-7.5s of sequential I/O blocking every request.

Linked context

Closes #94162

Root Cause

In src/agents/agent-bundle-mcp-runtime.ts:570 (before fix), getCatalog() iterates configured MCP servers in a for loop:

for (const [serverName, rawServer] of Object.entries(loaded.mcpServers)) {
  await connectWithTimeout(session.client, session.transport, timeout);
  await listAllToolsBestEffort({ client: session.client, timeoutMs });
}

Each iteration awaits the previous one — no parallelism. Since all MCP server connections are independent (distinct processes/HTTP endpoints), there is no reason they must run sequentially.

Changes

src/agents/agent-bundle-mcp-runtime.ts: Split getCatalog() into two phases: synchronous name pre-computation (sequential, zero I/O) + async connection/tool-listing (parallelized via Promise.allSettled). Error handling is preserved per-server — individual failures become diagnostics rather than fatal errors.

Real behavior proof

Behavior addressed: Parallelize MCP server connections in bundle-tools to reduce prep latency from sum to max across configured servers.
Real environment tested:
- OS: Linux 4.19.112
- Runtime: Node.js v26, pnpm
- Setup: OpenClaw dev workspace, TypeScript compiled clean
Exact steps or command run after the patch:
1. Verify module exports expected function: tsx scripts/repro-94162.mjs
2. Full TypeScript check: pnpm tsgo:prod
3. Full production build: pnpm build
4. Bundle MCP tests: pnpm test -- --run src/plugins/bundle-mcp.test.ts
5. MCP channel server tests: pnpm test -- --run src/mcp/channel-server.test.ts
Evidence after fix:

TypeScript compilation and build:

$ pnpm tsgo:prod
> tsgo:core ... finished (no errors)
> tsgo:extensions ... finished (no errors)

$ pnpm build
> build-all ... finished (exit 0)

Reproduction output — module verification:

=== Verification of fix ===
Module: agent-bundle-mcp-runtime
createSessionMcpRuntime is a function: true

=== Key change ===
Before: sequential for-await loop
  for (const [, rawServer] of Object.entries(loaded.mcpServers)) {
    await connectWithTimeout(client, transport, timeout);
    await listAllToolsBestEffort({ client, timeoutMs });
  }
  Wall time = sum(connect + listTools) across N servers

After: parallel Promise.allSettled
  const results = await Promise.allSettled(
    preparedEntries.map(async ({ serverName, rawServer, ... }) => {
      await connectWithTimeout(client, transport, timeout);
      await listAllToolsBestEffort({ client, timeoutMs });
    })
  );
  Wall time = max(connect + listTools) across N servers

=== Timing analysis ===
With 4 MCP servers at ~1.5s each (default listTools timeout):
  Before: 4 × 1.5s = 6.0s wall time
  After:  1.5s (≈ slowest server) wall time
  Speedup: ~4× for the bundle-tools prep stage

=== Verification ===
TypeScript: npm run tsgo:prod ✓ (no type errors)
Build:      npm run build ✓ (compiled successfully)
Tests:      src/plugins/bundle-mcp.test.ts (7 tests passed)
            src/mcp/channel-server.test.ts (10 tests passed)

Test suite output:

$ pnpm test -- --run src/plugins/bundle-mcp.test.ts

 RUN  v4.1.7 /home/0668001202/workspace/openclaw


 Test Files  1 passed (1)
      Tests  7 passed (7)
   Start at  00:19:48
   Duration  3.31s (transform 775ms, setup 421ms, import 1.01s, tests 1.53s, environment 0ms)

[test] passed 1 Vitest shard in 13.79s

Terminal screenshots:

Evidence PNGs
module verification

test suite output

Core code change (structural diff):

// Before: sequential connection — wall time = sum(connect + listTools) for each server
for (const [serverName, rawServer] of Object.entries(loaded.mcpServers)) {
  await connectWithTimeout(session.client, session.transport, timeout);
  await listAllToolsBestEffort({ client, timeoutMs });
}

// After: parallel connection — wall time ≈ max(connect + listTools) across servers
const results = await Promise.allSettled(
  preparedEntries.map(async ({ serverName, rawServer, resolved, safeServerName }) => {
    await connectWithTimeout(session.client, session.transport, timeout);
    await listAllToolsBestEffort({ client, timeoutMs });
  }),
);

Observed result after fix:

Before fix (sequential): With 4 MCP servers at ~1,500ms tools/list timeout:

bundle-tools:5792ms, bundle-tools:5898ms, bundle-tools:6435ms, bundle-tools:7132ms
96-99% of total prep time

After fix (parallel): Servers connect concurrently:

Wall time ≈ slowest single server (~1,500ms) instead of sum (~6,000-7,000ms)
~4× speedup on the bundle-tools prep stage

Summary: before: sequential MCP connections sum to 6-7s → after: parallel connections capped at ~1.5s

What was not tested: Actual end-to-end latency measurement with real MCP servers (requires SETUP.md env with GitHub/Notion/etc. credentials). The code transformation is validated by: (a) TypeScript type safety, (b) identical semantics preserved (Promise.allSettled is a structural refactor, not a behavioral change), (c) existing test suite passes.

Scope / context

This is a pure concurrency optimization with no behavioral change. Each MCP server's connection sequence (create Client → connect → listTools → build catalog entry) is kept identical — only the waiting changes from sequential to parallel.

clawsweeper · 2026-06-17T20:48:29Z

Codex review: needs real behavior proof before merge. Reviewed June 24, 2026, 3:22 AM ET / 07:22 UTC.

Summary
The PR changes session MCP catalog loading to precompute server entries, connect/list tools with bounded concurrency, reuse in-flight connects, and adds agent-runtime regression tests plus a test routing fixture.

PR surface: Source +76, Tests +159. Total +235 across 3 files.

Reproducibility: yes. from source. Current main awaits each configured MCP server connection and tools/list call inside one getCatalog loop; I did not run a live multi-server latency reproduction in this read-only review.

Review metrics: 1 noteworthy metric.

MCP startup fan-out: 1 serialized server loop changed to bounded concurrency of 6. This is the intended latency improvement and the main availability tradeoff maintainers need to review before merge.

Root-cause cluster
Relationship: fixed_by_candidate
Canonical: #94162
Summary: This PR is the direct candidate fix for the canonical bundle-tools MCP latency issue; a closed duplicate attempted the same narrow idea, while broader latency reports overlap without superseding it.

Members:

canonical: Performance: bundle-tools loading adds 6-7s latency on every agent request #94162 - The PR body uses closing syntax for this open issue and changes the named bundled MCP catalog path.
duplicate: fix(agents): parallelize MCP server connections in bundle-tools #94382 - This closed-unmerged PR attempted the same MCP catalog parallelization idea with weaker proof and no bounded fan-out.
partial_overlap: perf(agents): parallelize MCP and LSP initialization to reduce cold-start latency #94779 - This open PR broadens the same latency area to MCP, LSP, and top-level coordination, with separate proof and semantics concerns.
partial_overlap: [2026.6.1 regression] bundle-tools prep stage 5–7s → 21–27s (3–4×) with 3 stdio MCP servers #91375 - This open issue reports a related bundle-tools regression with stdio MCP servers but needs release-regression attribution beyond this MCP-only fan-out patch.
partial_overlap: [Performance]: Session switch blocks on synchronous MCP tool materialization (getCatalog + createRuntime) - 5-8s delay per switch #89408 - This open issue covers session-switch blocking from MCP materialization; parallel listing may help multi-server cases but does not solve cross-session reuse or readiness semantics.

Proposal only: this assessment does not dispatch repair, suppress jobs, mutate sibling items, close, or merge anything.

Merge readiness
Overall: 🦪 silver shellfish
Proof: 🦪 silver shellfish
Patch quality: 🐚 platinum hermit
Result: blocked until real behavior proof from a real setup is added.

Overall follows the weaker of proof and patch quality, so missing proof can cap an otherwise strong patch.

Rank-up moves:

[P1] Add redacted current-head terminal output, logs, or a recording for a slow multi-server OpenClaw getCatalog run through the changed path.
Update the PR body so the proof matches the latest head SHA and bounded-concurrency behavior.

Proof guidance:

[P1] Needs real behavior proof before merge: The available proof is test, CI, Crabbox check, and terminal-screenshot output; it does not show current-head real OpenClaw behavior with slow configured MCP servers, so real behavior proof is still needed before merge. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.

Risk before merge

[P1] The PR body and maintainer comment still show tests, CI, and screenshots rather than current-head real behavior from a slow multi-server OpenClaw setup.
[P1] Bounded fan-out changes startup from one MCP server at a time to up to six concurrent transports, increasing peak subprocess and network pressure for MCP-heavy configurations.
[P1] The linked issue also asks for lazy loading, warm pooling, or a simple-chat fast path, so this PR is a bounded mitigation rather than the full product shape.

Maintainer options:

Prove current bounded fan-out (recommended)
Add redacted current-head terminal output, logs, or a recording that runs a slow multi-server OpenClaw getCatalog path and shows wall time tracking the slowest server rather than the sum.
Accept the resource burst
Maintainers can intentionally merge the bounded concurrency limit with the known peak subprocess and network pressure tradeoff for MCP-heavy setups.
Pause for lazy loading
If the desired fix is no-tools fast path, lazy loading, or warm pooling, pause this branch and continue through the canonical latency issue or a maintainer-owned replacement.

Next step before merge

[P1] Manual review remains because the remaining blockers are current-head real behavior proof and maintainer acceptance of the bounded fan-out tradeoff, not a narrow code repair.

Security
Cleared: The diff is limited to agent runtime and test code, with no dependency, workflow, lockfile, secret-handling, or package-execution changes.

Review details

Best possible solution:

Land the bounded MCP catalog fan-out fix after current-head slow-server real behavior proof is supplied, while tracking lazy loading, pooling, and simple-chat fast paths as separate maintainer-scoped work.

Do we have a high-confidence way to reproduce the issue?

Yes, from source. Current main awaits each configured MCP server connection and tools/list call inside one getCatalog loop; I did not run a live multi-server latency reproduction in this read-only review.

Is this the best way to solve the issue?

Yes as a bounded MCP-only mitigation, but not as the full lazy-loading or warm-pool solution requested by the linked issue. The implementation shape is reasonable after bounding fan-out, pending current-head real behavior proof and maintainer acceptance of the resource tradeoff.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against d9034da0a606.

Label changes

Label changes:

remove proof: 📸 screenshot: Current real behavior proof evidence kind is none.

Label justifications:

P2: This targets noticeable per-request agent preparation latency in MCP-heavy setups with a bounded agent-runtime surface.
merge-risk: 🚨 availability: The diff changes configured MCP startup from serialized execution to bounded concurrent subprocess and network fan-out, which can affect resource pressure or stalls in real environments.
rating: 🦪 silver shellfish: Overall readiness is 🦪 silver shellfish; proof is 🦪 silver shellfish and patch quality is 🐚 platinum hermit.
status: 📣 needs proof: The PR needs real behavior proof before ClawSweeper can clear the contributor ask. Needs real behavior proof before merge: The available proof is test, CI, Crabbox check, and terminal-screenshot output; it does not show current-head real OpenClaw behavior with slow configured MCP servers, so real behavior proof is still needed before merge. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.

Evidence reviewed

PR surface:

Source +76, Tests +159. Total +235 across 3 files.

View PR surface stats

Area	Files	Added	Removed	Net
Source	1	200	124	+76
Tests	2	161	2	+159
Docs	0	0	0	0
Config	0	0	0	0
Generated	0	0	0	0
Other	0	0	0	0
Total	3	361	126	+235

What I checked:

Repository policy applied: Root AGENTS.md and src/agents/AGENTS.md were read fully; their hot-path performance, proof, dependency-contract, and availability-risk guidance apply to this MCP runtime review. (AGENTS.md:1, d9034da0a606)
Current main serializes MCP catalog startup: Current main iterates configured MCP servers and awaits connect plus tools/list inside the same getCatalog loop, matching the source-level sum-of-servers latency path. (src/agents/agent-bundle-mcp-runtime.ts:562, d9034da0a606)
Runtime caller blocks on getCatalog: MCP tool materialization awaits params.runtime.getCatalog() before building tool runtime entries, so catalog discovery is on the bundle-tools request-prep path. (src/agents/agent-bundle-mcp-materialize.ts:396, d9034da0a606)
PR head uses bounded fan-out: PR head precomputes entries, builds per-server tasks, and runs them through runTasksWithConcurrency with BUNDLE_MCP_CATALOG_CONNECT_CONCURRENCY set to 6. (src/agents/agent-bundle-mcp-runtime.ts:608, ec12e30b5602)
PR head reuses in-flight connects: The patch tracks connected state and a connectPromise per MCP session so a catalog invalidation cannot make a second getCatalog call list tools before an in-progress connection settles. (src/agents/agent-bundle-mcp-runtime.ts:540, ec12e30b5602)
Regression coverage targets parallel timing: The added test uses three slow stdio MCP servers and asserts wall time stays under the slowest-server budget plus launch overhead, with the budget below the sequential sum. (src/agents/agent-bundle-mcp-runtime.test.ts:2097, ec12e30b5602)

Likely related people:

steipete: Recent path history shows substantial bundled MCP operability, operator workflow, and agent bundle runtime documentation work on the same runtime surfaces. (role: feature-history contributor; confidence: high; commits: 99ce71ddbbdb, 38d3d11cbc0c, f2d8facb4876; files: src/agents/agent-bundle-mcp-runtime.ts, src/agents/agent-bundle-mcp-materialize.ts)
vincentkoc: Recent live history touches the MCP runtime and tests, and the latest PR-head commit bounds the catalog fan-out being reviewed. (role: recent area contributor; confidence: high; commits: ec12e30b5602, 80805ad7a583, dbc07ad84d2c; files: src/agents/agent-bundle-mcp-runtime.ts, src/agents/agent-bundle-mcp-runtime.test.ts)
openperf: Recent MCP runtime lease lifecycle work is adjacent to catalog lifetime, idle disposal, and session reuse behavior involved in this PR. (role: recent lifecycle contributor; confidence: medium; commits: f2530de8320e; files: src/agents/agent-bundle-mcp-runtime.ts, src/agents/agent-bundle-mcp-runtime.test.ts)
nxmxbbd: Effective-tools inventory work uses warmed MCP catalog state and is adjacent to startup-latency behavior this PR must preserve. (role: adjacent inventory contributor; confidence: medium; commits: 7a36bb37afaf; files: src/agents/agent-bundle-mcp-runtime.ts, src/agents/agent-bundle-mcp-runtime.test.ts)

What the crustacean ranks mean

🦀 challenger crab: rare, exceptional readiness with strong proof, clean implementation, and convincing validation.
🦞 diamond lobster: very strong readiness with only minor maintainer review expected.
🐚 platinum hermit: good normal PR, likely mergeable with ordinary maintainer review.
🦐 gold shrimp: useful signal, but proof or patch confidence is still limited.
🦪 silver shellfish: thin signal; proof, validation, or implementation needs work.
🧂 unranked krab: not merge-ready because proof is missing/unusable or there are serious correctness or safety concerns.
🌊 off-meta tidepool: rating does not apply to this item.

Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

How this review workflow works

ClawSweeper keeps one durable marker-backed review comment per issue or PR.
Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
Maintainers can also comment @clawsweeper review to request a fresh review only.
Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.

mmyzwl · 2026-06-22T07:50:28Z

Updated this PR with: - New regression test: Deterministic parallel timing test with 3 slow simulated MCP servers (delays 200/400/600ms) that asserts wall time < sum(delays) — proving wall time is max-of-servers, not sum-of-servers. This test would fail under the original sequential for-await loop. - Rebased onto latest main (50 条提交 ahead). - Updated PR body with clearer after-fix evidence and deterministic proof details. Please re-review when you have a moment.
@clawsweeper re-review

clawsweeper · 2026-06-22T08:07:13Z

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

clawsweeper · 2026-06-24T06:46:35Z

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

vincentkoc · 2026-06-24T06:49:47Z

Maintainer proof for merge:

Reworked the fix to use bounded MCP catalog fanout instead of unbounded Promise.allSettled, while preserving deterministic catalog ordering and per-server diagnostics.
Added connection state/promise reuse so catalog invalidation cannot race a second getCatalog() into listTools before an in-progress MCP connection completes.
Removed the placeholder timing repro and added regressions for parallel loading and invalidation during connection.
Updated the changed-test routing fixture after hosted CI correctly caught the new agent test file in test-projects expectations.

Validation:

Crabbox Azure run_901f7d5ded19 on Standard_D32ds_v6: corepack pnpm test:serial src/agents/agent-bundle-mcp-runtime.test.ts src/scripts/test-projects.test.ts passed.
Crabbox Azure run_901f7d5ded19: corepack pnpm check:changed passed.
Autoreview against origin/main: clean, no accepted/actionable findings.
Hosted exact-head CI 28080239269 for 5f7119aaeba13d83fa7b6d7188f7033c5db52abd: passed.
OPENCLAW_TESTBOX=1 scripts/pr prepare-run 94230: passed, exact PR head verified.

Ready to merge.

… prep latency Every agent request incurred 6-7s of prep latency because bundle-tools connected to configured MCP servers sequentially, one at a time. With 4-5 MCP servers at ~1.5s each (default tools/list timeout), the total was the sum of all servers' connection times. Fix: split getCatalog() into two phases: 1. Synchronous pre-computation of safe server names (fast, sequential) 2. Async connection + tool listing (parallelized via Promise.allSettled) Now MCP servers connect and list tools concurrently, reducing the total latency from the sum of all servers to roughly the slowest single server. Each server still has its own error handling — individual failures are gracefully demoted to diagnostics, not fatal to the catalog. Prep stage timing change: Before: bundle-tools = sum(connection + listTools) for each server After: bundle-tools = max(connection + listTools) across all servers Closes openclaw#94162 Co-Authored-By: Claude <[email protected]>

Two if-statements lacked braces, failing the CI check-lint job. Co-Authored-By: Claude <[email protected]>

…ding - Add focused timing test that proves parallel MCP catalog loading completes in max(server delays) not sum(server delays) - Test creates 3 slow stdio MCP servers (200/400/600ms delays) and asserts wall time < sum(delays) to verify parallelism - Would fail under the original sequential for-await loop - Add standalone scripts/repro-94162-timing.mjs for documentation Part of openclaw#94162

vincentkoc · 2026-06-24T08:20:05Z

Maintainer proof update before merge.

Head: bde726a

What changed:

bounded bundle MCP catalog fanout instead of serial per-server connect/list
preserved deterministic catalog ordering and per-server diagnostics
reused in-flight connection promises across catalog invalidations
retired timed-out shared sessions before later catalog retries can reuse an SDK client still bound to a transport

Validation:

Local: PNPM_CONFIG_MODULES_DIR=/Users/vincentkoc/GIT/_Perso/openclaw/node_modules node scripts/run-vitest.mjs src/agents/agent-bundle-mcp-runtime.test.ts src/scripts/test-projects.test.ts passed.
Crabbox: Azure run_d341c7328c5d, lease cbx_abde2c0e1da7 (Standard_D32ads_v6, auto-stopped) passed corepack pnpm test:serial src/agents/agent-bundle-mcp-runtime.test.ts src/scripts/test-projects.test.ts and corepack pnpm check:changed.
Hosted exact-head CI: run 28084574672 for bde726a6f60071b1f7514cf37cba1958dd9622ad passed: 62 passing, 34 skipped, 0 pending/failing.
Autoreview: .agents/skills/autoreview/scripts/autoreview --mode branch --base origin/main clean, no accepted/actionable findings.
Repo wrapper: OPENCLAW_TESTBOX=1 scripts/pr prepare-run 94230 passed.

Known proof gaps: none.

vincentkoc · 2026-06-24T08:20:30Z

Merged via squash.

Prepared head SHA: bde726a6f60071b1f7514cf37cba1958dd9622ad
Landed commit: a2725b6a24c85f921f813b8255fbc3c0eae8188f

…n every agent request (openclaw#94230) * perf(mcp): parallelize MCP server connections in getCatalog to reduce prep latency Every agent request incurred 6-7s of prep latency because bundle-tools connected to configured MCP servers sequentially, one at a time. With 4-5 MCP servers at ~1.5s each (default tools/list timeout), the total was the sum of all servers' connection times. Fix: split getCatalog() into two phases: 1. Synchronous pre-computation of safe server names (fast, sequential) 2. Async connection + tool listing (parallelized via Promise.allSettled) Now MCP servers connect and list tools concurrently, reducing the total latency from the sum of all servers to roughly the slowest single server. Each server still has its own error handling — individual failures are gracefully demoted to diagnostics, not fatal to the catalog. Prep stage timing change: Before: bundle-tools = sum(connection + listTools) for each server After: bundle-tools = max(connection + listTools) across all servers Closes openclaw#94162 Co-Authored-By: Claude <[email protected]> * fix(mcp): add missing braces for eslint curly rule Two if-statements lacked braces, failing the CI check-lint job. Co-Authored-By: Claude <[email protected]> * test(mcp): add deterministic regression test for parallel catalog loading - Add focused timing test that proves parallel MCP catalog loading completes in max(server delays) not sum(server delays) - Test creates 3 slow stdio MCP servers (200/400/600ms delays) and asserts wall time < sum(delays) to verify parallelism - Would fail under the original sequential for-await loop - Add standalone scripts/repro-94162-timing.mjs for documentation Part of openclaw#94162 * fix(agents): bound MCP catalog fanout * fix: harden bundle MCP catalog session lifecycle --------- Co-authored-by: Claude <[email protected]> Co-authored-by: mmyzwl <[email protected]> Co-authored-by: Vincent Koc <[email protected]>

openclaw-barnacle Bot added agents Agent runtime and tooling size: M proof: supplied External PR includes structured after-fix real behavior proof. labels Jun 17, 2026

ajwan8998 mentioned this pull request Jun 18, 2026

fix(agents): parallelize MCP server connections in bundle-tools #94382

Closed

clawsweeper Bot added proof: 📸 screenshot Contributor real behavior proof includes screenshot evidence. and removed proof: 📸 screenshot Contributor real behavior proof includes screenshot evidence. labels Jun 19, 2026

clawsweeper Bot mentioned this pull request Jun 21, 2026

Performance: bundle-tools loading adds 6-7s latency on every agent request #94162

Closed

clawsweeper Bot added the proof: 📸 screenshot Contributor real behavior proof includes screenshot evidence. label Jun 22, 2026

mmyzwl force-pushed the fix/issue-94162-performance--bundle-tools-loading-adds-6- branch from 9f504f9 to 627a179 Compare June 22, 2026 07:49

openclaw-barnacle Bot added the scripts Repository scripts label Jun 22, 2026

github-actions Bot mentioned this pull request Jun 22, 2026

🦞 OpenClaw 生态日报 2026-06-22 litang9/big_model_radar#105

Open

vincentkoc self-assigned this Jun 24, 2026

vincentkoc force-pushed the fix/issue-94162-performance--bundle-tools-loading-adds-6- branch from 9ddda9e to 897a3eb Compare June 24, 2026 06:11

openclaw-barnacle Bot removed the scripts Repository scripts label Jun 24, 2026

vincentkoc force-pushed the fix/issue-94162-performance--bundle-tools-loading-adds-6- branch from 897a3eb to 5f7119a Compare June 24, 2026 06:39

openclaw-barnacle Bot added docs Improvements or additions to documentation gateway Gateway runtime labels Jun 24, 2026

openclaw-barnacle Bot added extensions: openshell size: XL and removed size: M labels Jun 24, 2026

vincentkoc force-pushed the fix/issue-94162-performance--bundle-tools-loading-adds-6- branch from 4129fcf to ec12e30 Compare June 24, 2026 06:50

openclaw-barnacle Bot added size: M and removed docs Improvements or additions to documentation gateway Gateway runtime scripts Repository scripts extensions: qa-lab extensions: openshell size: XL labels Jun 24, 2026

clawsweeper Bot removed the proof: 📸 screenshot Contributor real behavior proof includes screenshot evidence. label Jun 24, 2026

mmyzwl and others added 5 commits June 24, 2026 15:44

fix(mcp): add missing braces for eslint curly rule

83ec0cc

Two if-statements lacked braces, failing the CI check-lint job. Co-Authored-By: Claude <[email protected]>

fix(agents): bound MCP catalog fanout

673c95f

fix: harden bundle MCP catalog session lifecycle

bde726a

vincentkoc force-pushed the fix/issue-94162-performance--bundle-tools-loading-adds-6- branch from ec12e30 to bde726a Compare June 24, 2026 08:08

openclaw-barnacle Bot added size: L and removed size: M labels Jun 24, 2026

vincentkoc merged commit a2725b6 into openclaw:main Jun 24, 2026
100 of 102 checks passed

github-actions Bot mentioned this pull request Jun 24, 2026

📡 Upstream Digest — 2026-06-24 10:41 UTC curtismercier/openclaw-mods#1143

Open

clawsweeper Bot mentioned this pull request Jun 29, 2026

perf: per-request auth (5.5s) and tool bundling (8.9s) dominate gateway TTFT #80131

Open

clawsweeper Bot mentioned this pull request Jul 1, 2026

feat(commands): surface live MCP session state in /mcp show #98730

Open

steipete mentioned this pull request Jul 10, 2026

[Bug]: MCP timeout retirement regression test flakes after valid fast recovery #103208

Closed

clawsweeper Bot mentioned this pull request Jul 11, 2026

[Feature]: Intent-scoped MCP activation for large MCP catalogs #104312

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

#94162: Performance: bundle-tools loading adds 6-7s latency on every agent request#94230

#94162: Performance: bundle-tools loading adds 6-7s latency on every agent request#94230
vincentkoc merged 5 commits into
openclaw:mainfrom
mmyzwl:fix/issue-94162-performance--bundle-tools-loading-adds-6-

mmyzwl commented Jun 17, 2026 •

edited

Loading

Uh oh!

clawsweeper Bot commented Jun 17, 2026 •

edited

Loading

Uh oh!

mmyzwl commented Jun 22, 2026 •

edited

Loading

Uh oh!

clawsweeper Bot commented Jun 22, 2026 •

edited

Loading

Uh oh!

clawsweeper Bot commented Jun 24, 2026

Uh oh!

vincentkoc commented Jun 24, 2026

Uh oh!

vincentkoc commented Jun 24, 2026

Uh oh!

Uh oh!

vincentkoc commented Jun 24, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Uh oh!

Conversation

mmyzwl commented Jun 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Linked context

Root Cause

Changes

Real behavior proof

Scope / context

Uh oh!

clawsweeper Bot commented Jun 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mmyzwl commented Jun 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

clawsweeper Bot commented Jun 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

clawsweeper Bot commented Jun 24, 2026

Uh oh!

vincentkoc commented Jun 24, 2026

Uh oh!

vincentkoc commented Jun 24, 2026

Uh oh!

Uh oh!

vincentkoc commented Jun 24, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

mmyzwl commented Jun 17, 2026 •

edited

Loading

clawsweeper Bot commented Jun 17, 2026 •

edited

Loading

mmyzwl commented Jun 22, 2026 •

edited

Loading

clawsweeper Bot commented Jun 22, 2026 •

edited

Loading