EdgeBuddy is the epitomy of sophisticated backend to "one click sync" to have access to a trading companion that connects the dots of your trading performance in ways that matter.
Finally, someone who actually "gets" you 24/7.
@nummanali You can literally SSH over tailscale and use any cli session you want. Still preachimg vendor locked as the solution shouldnt be the norm. (From a person working mainly on codex cli) if a frontier model changes and is more helpful or i need other alternatives my stack adapts.
@nummanali Sth sth... agnostic architecture vs vendor locked. Thanks for the attention to the matter. Never ever stay locked in one architecture where you are dependent to the vendor
INTEL: Agentic AI costs are blowing past enterprise budgets, with Microsoft scaling back Claude Code access and Uber burning through its 2026 AI coding budget in four months
This GPT-Realtime-2 bridge to your OpenClaw agents is now ready for wider use. Native iOS app. Install the free TestFlight app on iOS and DM me your email address to start using it immediately via a TestFlight “internal tester” invitation. Otherwise, the public TestFlight “external tester” link will be available in a few hours.
The macOS companion app is linked in the first reply below.
I may change the app name later, but for now the real work of getting it publicly available in a native iOS app is done. Let me know what you think!
@pashmerepat@bitcoinduke So the know compat for compatibility with openclaw tooling is gone? Until v5.7 i was using codex harness with openclaw compatibility to make tooling functional! Thanks for your contribution
Thanks, that clarifies the diagnostic boundary!
I’ll stop treating durable task creation as expected for normal WhatsApp inbound turns. The remaining issue is still after inbound/group gating/session metadata: no visible model/run/auto-reply/delivery-suppressed logs
Redacted split:
accountId: default
sessionKey: agent:<agent-id>:whatsapp:group:<group-jid>
Inbound:
2026-05-13T09:26:57.879+02:00 [whatsapp] Inbound message <group-jid> ->
<agent-phone> (group, 105 chars)
Then:
2026-05-13T09:29:14.483+02:00 [diagnostic] stuck session:
sessionKey=agent:<agent-id>:whatsapp:group:<group-jid> state=processing
age=137s queueDepth=1 reason=queued_work_without_active_run
classification=stale_session_state recovery=checking
2026-05-13T09:29:14.491+02:00 [diagnostic] stuck session recovery no-op:
reason=no_active_work action=none sessionKey=agent:<agent-
id>:whatsapp:group:<group-jid>
I don’t see model/run activity, “Skipping auto-reply”, “Delivery suppressed”,
or “auto-reply delivery failed” around that inbound.
Apologies for not filing it on GH for now
OpenClaw v2026.5.7 WhatsApp group inbound regression help wanted.
Clean macOS rebuild, fresh WhatsApp link, simple route:
whatsapp/default -> agent
Expected:
allowed group message -> agent run/task -> visible group reply
Actual:
WhatsApp receives the group message, channel status updates, group session
metadata is created, but no agent task/run is created and no outbound WhatsApp
send happens.
Pipeline:
WhatsApp inbound ��
group allowlist / mention gating ✅
session metadata ✅
runInboundReplyTurn / channel dispatch ❌
durable task creation ❌
Config shape:
- DMs disabled
- groupPolicy=allowlist
- one allowlisted WhatsApp group
- requireMention=false
- groupAllowFrom=["*"]
- messages.visibleReplies=automatic
- messages.groupChat.visibleReplies=automatic
- top-level route binding: whatsapp/default -> agent
Also tested the upstream no_active_work stuck-session recovery fix from PR
#77865. It applied cleanly, but did not change the symptom, so this seems
earlier than recovery/cleanup: the inbound turn never becomes agent work.
Question:
Has anyone seen v2026.5.7 silently stop between WhatsApp process-message.ts /
runInboundReplyTurn and agent task creation?
@steipete @openclaw
@witcheer yes ! basically just gives my my cli straight in my phone provided you have also tailscale for ssh. maybe yours do other things as well but termius for cli is awesome cause even when am at my mac am only though codex cli no codex app for me
Codex multi-agent v2 + goals is the interesting part.
Goals give the session a durable mission:
- objective
- scope
- acceptance criteria
- verification
- stop conditions
Multi-agent v2 gives the main agent a live team:
- spawn specialists
- list active agents
- steer them mid-task
- assign follow-ups
- close stale/finished threads
The combo changes the workflow.