Skip to content

[P0][Bug]: Native Store Gateway onboarding requires manual clawctl setup before the wizard can continue #1672

Description

@joelagnel

Bug

Requested priority: P0. Native Store onboarding blocks users until they run an external repair command.

Microsoft Store Companion native Gateway onboarding stops at Prepare your Gateway profile with an unsupported isolated-session contract error. Running the installed Store Gateway's clawctl setup --json manually, then clicking Retry setup, allows the wizard to continue using the same installed package versions.

The wizard should prepare the supported Gateway automatically. A user should not need to leave onboarding and run a terminal command.

Reproduction

Observed once on October 7, 2026:

  1. Uninstall the existing developer Companion package and back up the previous Companion profile.
  2. Install Companion from Microsoft Store, product 9NFPR3BGDRR5, and launch with a fresh Companion profile.
  3. Select Install a local native gateway, the recommended option on this host.
  4. Let the wizard install Gateway from Microsoft Store, product 9NV70LV3D6XC.
  5. Observe that Check device support and Install or verify Gateway package complete, but Prepare your Gateway profile shows Needs attention. Start and verify the Gateway remains waiting.

This was an existing development machine with a fresh Companion profile. Gateway/session state was not deliberately reset before the test. A factory-clean machine and a current-main build have not been tested for this failure.

Expected behavior

The wizard initializes or safely repairs the supported native Gateway and continues without a manual clawctl setup invocation. If preparation cannot proceed, it surfaces the actual readiness problem with an actionable recovery path.

Actual behavior

The wizard displays:

The installed Gateway MSIX does not expose Companion's isolated-session contract. Update the Gateway MSIX, then retry setup. Credentials were not sent.

The installed Gateway CLI, invoked in the signed-in Windows user's interactive session, reported that the isolated-session helper had not been staged for this package version and directed the user to run clawctl setup again.

Environment

Component Observed version
Companion MSIX OpenClawFoundation.OpenClaw, 2026.9.520.0, ARM64
Gateway MSIX OpenClawFoundation.OpenClawGateway, 2026.9.701.0, ARM64
Package provenance Both SignatureKind=Store, Status=Ok
Windows Windows 11 Enterprise Insider Preview, 26H1, ARM64, 28120.3151
GPU NVIDIA RTX Spark N1X, 6144-core Blackwell RTX GPU
NVIDIA driver 616.00

Companion package family: OpenClawFoundation.OpenClaw_rfcbke2p71se2.
Gateway package family: OpenClawFoundation.OpenClawGateway_rfcbke2p71se2.

The wizard's native device-support check passed on this Windows build. The observed failure occurred later during Gateway profile preparation, and the workaround succeeded without a Windows update.

No source build, developer MSIX replacement, or WSL fallback was used for the successful run.

Workaround and observed recovery

Ran the installed Store Gateway's package-qualified alias with setup --json in the signed-in user's interactive Windows session. Equivalent PowerShell:

$package = Get-AppxPackage -Name OpenClawFoundation.OpenClawGateway
$clawctl = Join-Path $env:LOCALAPPDATA ('Microsoft\WindowsApps\' + $package.PackageFamilyName + '\clawctl.exe')
& $clawctl setup --json

The command reported success, a ready isolated session, and integration isolated-session version 1. Clicking Retry setup then allowed native onboarding to continue. Neither MSIX was updated between failure and recovery.

Completed Local AI installation through the wizard. It reached Your AI is ready with verified model llamacpp/qwen3.6-35b-a3b-mtp-ud-iq4-xs. After the setup-driven app restart, Chat showed Owner Connected, and Qwen 3.6 correctly answered 56 for 7 × 8. The app, native Gateway, and Local AI were left running.

Evidence

Original wizard failure:

Native Store onboarding fails at Gateway profile preparation

Wizard success after the manual command and retry:

Local AI wizard succeeds after native Gateway repair

Working local Chat after setup and restart:

Local Qwen 3.6 replies with the test nonce and 56 while Owner is connected

These are original screenshots from the observed run. A terminal screenshot of the repair command was not captured; the workaround above records the command that was invoked. The first failing package JSON response was not retained.

Investigation lead

The inspected source checks the package contract before preparing the package:

This ordering is consistent with an initialization/readiness failure stopping onboarding before automatic setup can repair it. It is an investigation lead, not a confirmed root cause. Regression coverage should include an installed supported package whose helper needs staging, and should preserve the existing credential-handoff checks.

Related merged work: #1553 (fix: Companion cannot onboard with an isolated Gateway), openclaw/openclaw-windows-packaging#134 (fix: Companion cannot configure an isolated Gateway), and #1649 (fix: Companion fails to reconnect after isolated session replacement). None was confirmed as an exact duplicate of this first-run failure.

Activity

  1. added
    P0Emergency: data loss, security bypass, crash loop, or unusable core runtime.
    clawsweeper:needs-live-reproClawSweeper needs live local, crabbox, or manual validation to confirm this issue.
    clawsweeper:needs-maintainer-reviewClawSweeper marked this issue as needing maintainer review before automation.
    clawsweeper:no-new-fix-prClawSweeper does not recommend queueing a new automated fix PR for this issue.
    impact:ux-release-blockerA non-technical user is blocked without terminal, logs, config, or support.
    issue-rating: 🐚 platinum hermitGood issue quality with a plausible reproduction path needing some confirmation.
    on Oct 7, 2026
  2. clawsweeper commented on Oct 7, 2026

    @clawsweeper

    Codex review: this still needs some work. Reviewed October 7, 2026, 10:01 PM ET / October 8, 2026, 02:01 UTC.

    Summary
    Keep open: the Store onboarding failure is documented, and the related merged packaging repair explicitly leaves this original incident unresolved.

    Reproducibility: no. the original failure occurred once, its failing JSON was lost, and the reporter’s clean same-version and missing-helper retests passed. A related controlled contention case was reproduced and repaired, but current-main reproduction of the original incident is not established.

    Next step
    Continue the retained-state investigation with redacted pre-repair package JSON; the accepted bounded repair leaves no confirmed, narrow Companion defect for automatic implementation.

    Review details

    Best possible solution:

    Preserve contract and credential checks, verify the original retained-state case with the packaging repair, and provide an inline Retry or Fix action for any remaining supported-package readiness failure.

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

    No: the original failure occurred once, its failing JSON was lost, and the reporter’s clean same-version and missing-helper retests passed. A related controlled contention case was reproduced and repaired, but current-main reproduction of the original incident is not established.

    Is this the best way to solve the issue?

    Unclear for the original incident: the package-owned error-envelope repair is appropriate for the demonstrated contention case, while changing Companion’s admission ordering without the original response would risk weakening established checks.

    AGENTS.md: found and applied where relevant.

    Remaining risk / open question:

    • The original pre-repair status JSON and retained-session trigger remain unknown; the merged contention repair does not establish resolution on the reported Store versions.

    Codex review notes: model internal, reasoning medium; reviewed against 037c17dcb581.

    Label changes

    Label changes:

    • add clawsweeper:no-new-fix-pr: Current issue advisory state selects this label.
    • add clawsweeper:needs-maintainer-review: Current issue advisory state selects this label.

    Label justifications:

    • P0: The documented Store onboarding block required an external terminal command before a first useful run.
    • impact:ux-release-blocker: Retry setup did not provide recovery until the user manually ran clawctl setup outside the wizard.
    Evidence reviewed

    What I checked:

    Likely related people:

    • paulcam206: Suggested for follow-up; no historical authorship or introduction is verified. (role: unverified routing candidate; confidence: low)
    • RomneyDa: Suggested for follow-up; no historical authorship or introduction is verified. (role: unverified routing candidate; confidence: low)
    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.
  3. joelagnel commented on Oct 7, 2026

    @joelagnel
    ContributorAuthor

    Clean Store retest

    Could not reproduce the original failure with a clean Store install. This is a diagnostic data point; it does not establish the original cause or a fix.

    Retested on Windows 11 Enterprise Insider Preview ARM64, build 28120.3151, using the same Store versions: Companion 2026.9.520.0 and Gateway 2026.9.701.0. This was native Gateway mode, not WSL.

    • Used supported Gateway teardown, backed up the prior Companion profile and managed Local AI files, removed both packages, and verified zero OpenClaw Appx registrations. Reinstalled Companion from Store and let its wizard install Gateway from Store. Existing WSL distributions were preserved; this was clean app/Gateway state on an existing OS.
    • Native support, package verification, profile preparation and Gateway startup all passed automatically. No external clawctl setup command was needed.
    • Separately stopped the new test Gateway, renamed its helper directory to preserve it, and restarted onboarding with another fresh Companion profile. The resulting helper-not-staged status already included integration.kind = isolated-session and version = 1. The unmodified Store wizard automatically restaged the helper and continued. This controlled case did not reproduce the contract error and does not support missing helper staging alone as its explanation.
    • Local AI setup passed. After the app restarted, Qwen 3.6 correctly answered 56 for 7 × 8, together with a new test nonce. Chat showed Owner Connected. The app, native Gateway and Local AI are running.

    No fix PR is being submitted because a same-issue fix is not yet confirmed. The original failure and successful manual workaround remain recorded above. Its first failing JSON response was not retained, so the actual root cause remains unconfirmed.

    Clean Local AI wizard completion:

    Clean native Store onboarding completes without an external setup command

    Working Chat after restart:

    Qwen 3.6 returns CLEAN_NATIVE_STORE_OK_20261007 56 with Owner Connected

  4. joelagnel commented on Oct 7, 2026

    @joelagnel
    ContributorAuthor

    I will leave this issue open for now since I am not sure if it has been fixed or if it is just not reproducible

  5. added
    status: 🚢 actively landingA maintainer or agent is actively driving this item through implementation, validation, or merge.
    on Oct 7, 2026
  6. added
    issue-rating: 🦪 silver shellfishThin issue quality; more reproduction proof or environment detail is needed.
    and removed
    issue-rating: 🐚 platinum hermitGood issue quality with a plausible reproduction path needing some confirmation.
    clawsweeper:needs-live-reproClawSweeper needs live local, crabbox, or manual validation to confirm this issue.
    on Oct 7, 2026
  7. removed
    status: 🚢 actively landingA maintainer or agent is actively driving this item through implementation, validation, or merge.
    on Oct 7, 2026
  8. removed
    clawsweeper:no-new-fix-prClawSweeper does not recommend queueing a new automated fix PR for this issue.
    clawsweeper:needs-maintainer-reviewClawSweeper marked this issue as needing maintainer review before automation.
    on Oct 7, 2026
  9. added
    status: 🚢 actively landingA maintainer or agent is actively driving this item through implementation, validation, or merge.
    and removed
    status: 🚢 actively landingA maintainer or agent is actively driving this item through implementation, validation, or merge.
    on Oct 7, 2026
  10. clawsweeper commented on Oct 7, 2026

    @clawsweeper

    🦞⚠️
    Automatic implementation stopped before completion.
    Reason: The automatic implementation worker stopped before all deterministic gates completed. Open the workflow run for the exact blocker.
    Worker: https://github.com/openclaw/clawsweeper/actions/runs/37714655762

  11. removed
    status: 🚢 actively landingA maintainer or agent is actively driving this item through implementation, validation, or merge.
    on Oct 8, 2026
  12. added
    clawsweeper:needs-maintainer-reviewClawSweeper marked this issue as needing maintainer review before automation.
    clawsweeper:no-new-fix-prClawSweeper does not recommend queueing a new automated fix PR for this issue.
    on Oct 8, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    P0Emergency: data loss, security bypass, crash loop, or unusable core runtime.clawsweeper:needs-maintainer-reviewClawSweeper marked this issue as needing maintainer review before automation.clawsweeper:no-new-fix-prClawSweeper does not recommend queueing a new automated fix PR for this issue.impact:ux-release-blockerA non-technical user is blocked without terminal, logs, config, or support.issue-rating: 🦪 silver shellfishThin issue quality; more reproduction proof or environment detail is needed.

    Type

    No type

    Fields

    Priority

    None yet

    Projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions