Problem

After upgrading from 2026.2.26 to 2026.3.1, every message in a Slack channel creates a separate agent context instead of continuing the same conversation.

Previously, chatting with OpenClaw in a Slack channel worked like a normal chatbot — each message in the channel shared the same agent session, so the agent remembered prior context. Now each message is completely isolated with no memory of previous messages.

My setup

• Using Slack channel conversations (not threads)
• replyToMode is off — I'm just chatting with OpenClaw like a normal LLM chatbot / Claude Code

Root cause

Commit 11d34700c ("fix(slack): use thread-level sessions for channels to prevent context mixing", #10686) changed the session key from:

agent:main:slack:channel:${channelId}

to:

agent:main:slack:channel:${channelId}:thread:${messageTs}

For users who chat in channels without threads, each top-level message gets its own ts, so every message creates a brand new isolated session. The agent has zero memory of anything said before.

Expected behavior

When replyToMode is off and I'm sending top-level messages in a channel, messages should share the same agent context — the same way it worked in 2026.2.26.

Suggested fix

The thread-level session isolation makes sense for threaded conversations, but there should be a way to opt out for users who use channels as a simple chat interface. Options:

1. Config flag like thread.sessionIsolation: false to keep channel-level sessions
2. Only apply thread-level sessions when the channel actually has active threads (i.e., when thread_ts differs from ts)
3. Fall back to channel-level session key when the message is a top-level message with no thread replies

Workaround

Downgraded to 2026.2.26.