slack-rate-limit-fix
OpenClaw + Slack: Fixing Context Loss After Slack's March 2026 Rate Limit Change
If your OpenClaw agent suddenly started giving vague, context-free responses in Slack — not remembering what you said a few minutes ago, missing thread context, or giving answers that feel disconnected from the conversation — Slack's March 3, 2026 API change is almost certainly the culprit.
This hit a lot of people without warning. Here's what happened, why it matters, and exactly how to fix it.
---
What Slack Changed
On March 3, 2026, Slack tightened the rate limits on conversations.history for apps not listed in the Slack Marketplace. The new limits are:
- 1 request per minute (down from effectively much higher)
- 15 messages maximum per call (down from 100)
For most personal OpenClaw setups, your agent uses a custom Slack app that isn't Marketplace-listed. That means it now hits these stricter limits every time it tries to read conversation history to understand context.
The result: instead of reading the last 50–100 messages to understand what's been discussed, your agent can now only read the last 15 — and slowly. In active channels or long threads, that means it's working with roughly 15% of the context it had before. Conversations feel like talking to someone with amnesia.
---
Why This Breaks Agents Specifically
Human Slack users aren't affected because they can see the whole conversation visually. Agents can't — they have to actively fetch message history on every response. When that fetch is throttled and capped, the agent's working context collapses.
Signs your setup is affected:
- Agent gives generic responses that don't reference recent messages
- Agent asks you to repeat information you already provided
- Agent loses track of multi-step tasks in a thread
- In busy channels, agent seems "confused" or off-topic
openclaw logsshows repeated 429 errors from the Slack API
---
Fix Option 1: Increase Local Context Cache (Recommended)
The most effective fix is to configure OpenClaw's Slack integration to cache messages locally rather than fetching from the API on every turn. This trades API calls for a persistent local store.
In your ~/.openclaw/openclaw.json:
`json
{
"integrations": {
"slack": {
"contextMode": "cache",
"cacheMessages": 200,
"cacheTTL": 3600
}
}
}
`
What this does: OpenClaw listens for new messages via the Events API (WebSocket, not rate-limited) and stores them in a local ring buffer. When your agent needs context, it reads from the local cache rather than calling conversations.history. You get full context, no rate limit issues.
Requirements: Your Slack app needs the messages:read Events API subscription enabled. If it's already posting and responding, this is likely already configured. Check your Slack app settings under "Event Subscriptions."
After updating the config, restart:
`bash
openclaw gateway restart
`
---
Fix Option 2: Reduce History Fetch Depth
If you'd rather not use local caching, you can reduce how aggressively OpenClaw fetches history — enough to stay under the 1 req/min limit without breaking context:
`json
{
"integrations": {
"slack": {
"contextMode": "fetch",
"maxHistoryMessages": 12,
"historyFetchInterval": 70
}
}
}
`
Setting maxHistoryMessages to 12 (under the new 15-message limit) avoids partial responses. Setting historyFetchInterval to 70 seconds keeps you safely under the 1/min rate limit.
This reduces context quality (12 messages vs. the previous 50–100), but it's stable and doesn't require any Slack app changes.
---
Fix Option 3: Get Your App Marketplace-Listed (Permanent Fix)
If you're running OpenClaw for a team or business, the permanent solution is getting your Slack app listed in the Marketplace. Marketplace apps get the standard (non-restricted) rate limits back.
This is a multi-week process that involves Slack's review, so it's not a quick fix — but worth pursuing if Slack is a critical integration for you. See Slack's developer documentation for the submission process.
---
Checking If the Fix Worked
After applying Fix Option 1 or 2 and restarting, send a few messages in Slack and then ask your agent something that requires recalling an earlier message:
`
Hey, what was the task I mentioned at the start of this conversation?
`
If it answers correctly, context is restored. You can also check openclaw logs --tail 50 — you should no longer see 429 errors from the Slack API.
---
Background: Why Non-Marketplace Apps?
Slack's rationale was that unapproved custom apps represented an unknown risk surface — they wanted to enforce tighter limits to reduce abuse potential while Marketplace-listed apps had gone through review. For personal productivity bots and team automations, this is a frustrating tradeoff, but the local cache workaround (Fix Option 1) largely neutralizes it.
---
Still Having Issues?
If you've tried the fixes above and Slack context is still broken:
openclaw doctor and look for Slack-related warningsopenclaw logs --tail 100 | grep -i slack for API errorsmessages:read event subscriptions enabled (for Fix 1)openclaw.json is valid: openclaw config --validate---
*Last updated: March 2026. Applies to OpenClaw 2026.2.x and later with Slack integration.*