OpenClaw Issue Digest: Azure OpenAI Blocks, ACP Regressions, and System Stability
The recent activity window for the OpenClaw repository reveals a mix of high-severity integration blockers, regressions in sub-agent orchestration, and several system-level stability issues. Most notably, a critical failure in the openai-responses adapter has rendered Azure OpenAI unusable for a significant subset of users, while regressions in the Agent Client Protocol (ACP) and Codex harness are complicating multi-agent workflows.
Open Issues
Integration & Model Failures
- Azure OpenAI Blockage (#79570): The
openai-responsesadapter is currently unusable against Azure OpenAI. Requests are short-circuited by Azure's prompt-stage shield, which classifies OpenClaw'sSender (untrusted metadata)envelope as an Indirect Prompt Injection, resulting in a synthetic 0-token refusal. Switching toopenai-completionsserves as a temporary workaround. - Anthropic Response Failures (#79492): Users report that
anthropic/claude-opus-4-7returns empty responses in the agent runtime, despite working in the lower-levelinfer model runpath. This is suspected to be caused by an unsupportedthinking.type=enabledrequest shape. - OpenRouter Media Gaps (#79535): Video generation via OpenRouter silently fails, and music generation is entirely missing from the provider registry.
Orchestration & ACP Regressions
- ACP Spawning Failures (#78546): Claude ACP sessions spawned via
sessions_spawnfail withACP_TURN_FAILED, even though the same setup works via the ACPX CLI. This blocks the use of Claude ACP as a managed coding sub-agent. - Codex Harness Eviction (#79495): In multi-agent setups, the Codex harness app-server shared client is evicted whenever the agent switches, because the cache key includes
agentDir. This leads to high CPU overhead and repeated reconnects on ARM hosts. - Sub-agent Task Visibility (#78592, #73656): Native
sessions_spawnhides the delegated task from the child's visible transcript, making it difficult to audit whether the child actually received the instructions or is simply running default bootstrap behavior.
System Stability & Performance
- Event Loop Blocking (#79264): Version 2026.5.7 introduced a regression where
post-attach.update-sentinelruns on every model call, causing ~9s synchronous blocks and triggeringevent_loop_delaywarnings. - Memory Growth & Freezes (#73691):
MEMORY.mdgrows unbounded when dreaming is enabled, eventually exceeding bootstrap limits and causing the Gateway to freeze due to session write-lock timeouts. - Android Notification Loss (#79552): The Android node sends
node.eventframes before the websocket handshake completes, leading to rejected notifications and broken automations. - Safe-FS Permission Errors (#79520): Outbound media uploads fail with a generic 500 error if files are mode
0o600, a result of the 2026.5.7 safe-FS rewrite interacting with hardcoded write permissions in the sandbox bridge.
Key Themes
The "Silent Failure" Pattern
Across multiple reports, a recurring theme is the lack of observability when a request is suppressed or fails. Examples include the visibleReplies="message_tool" config silently dropping text replies (#78405) and the openai-responses adapter returning synthetic refusals without log errors (#79570). Users are spending hours debugging infrastructure when the issue is a "working-as-designed" config gate or an upstream shield.
Multi-Agent Scaling Friction
As users move toward complex multi-agent orchestration, friction points are emerging in how state and resources are shared. The Codex harness singleton eviction (#79495) and the lack of per-agent isolation for inbound media (#19330) highlight the need for more granular resource pooling and scoping.
Packaging & Runtime Resolution
Several regressions point to issues with how dependencies are resolved after updates, specifically with the Matrix channel missing matrix-js-sdk (#77896) and Codex peer-link failures on Linux (#79462).
Action Required
Immediate Attention (High Severity)
- #79570 (Azure OpenAI): Critical blocker for Azure users. Requires a change to the
Senderenvelope or a config option to disable it. - #79264 (Event Loop Delay): Performance regression in 2026.5.7 that makes the system effectively unusable for some users.
- #79552 (Android Notifications): Core functionality loss for Android-based automation.
Blocked/Regression Fixes
- #78546 (Claude ACP): Needs investigation into why the OpenClaw spawn path differs from the ACPX CLI path.
- #79495 (Codex Harness): High priority for users on resource-constrained hardware (Raspberry Pi).
- #79520 (Safe-FS Permissions): Requires updating the sandbox bridge to use
0o644and improving error reporting for outbound media rejects.