Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.claw-link.dev/llms.txt

Use this file to discover all available pages before exploring further.

OpenClaw overview illustration This is where otherwise reasonable people accidentally build a tiny routing disaster. One bot becomes two. One agent becomes several. And suddenly everybody is online, nobody is predictable, and Discord starts feeling like a haunted switchboard. If you want to run multiple OpenClaw bots or agents in one Discord setup, the hard part is usually not getting another bot online. The hard part is making the routing and behavior understandable. Common problems in multi-bot or multi-agent Discord setups include:
  • bots respond in the wrong server, channel, or thread
  • one bot appears healthy while another silently absorbs the messages you meant to test
  • permissions differ between channels, making one bot look broken when it is really just under-scoped
  • multiple agents overlap in behavior, so users cannot tell who should respond to what
  • one setup works in DMs but collapses in shared channels
This guide is about keeping that setup intentional instead of accidental.
Monitor icon Useful mental model: in a multi-bot setup, the enemy is not usually installation. It is ambiguous ownership of message paths.

​
Lightbulb icon At a glance

  • The more bots or agents you add, the more important clear routing boundaries become.
  • Test one known route per bot first instead of dropping multiple bots into the same space and hoping for manners.
  • If behavior feels random, compare scope, permissions, and trigger rules before blaming the gateway.

​
What this problem usually looks like

Users usually mean one of these:
  • “We want one bot for support and another for internal ops.”
  • “We have multiple bots/agents in one Discord, and responses are inconsistent.”
  • “One bot works in some channels, another works elsewhere, and nobody remembers why.”
  • “Everything looked fine until we added the second bot.”
That is normal. A one-bot setup can get away with vagueness. A multi-bot setup usually cannot.

​
First: decide what each bot or agent owns

Before changing config, answer this clearly:
  • which bot/agent handles which channels?
  • which bot/agent handles which tasks?
  • which bot/agent is expected to respond in shared spaces?
  • do users need to mention a specific bot, or should behavior be automatic?
If you cannot answer those quickly, the setup is already too fuzzy.

​
Good patterns for multi-bot Discord setups

​
Pattern 1: split by channel ownership

Examples:
  • one bot for public support channels
  • one bot for internal staff channels
  • one bot for a private operations server
This is the cleanest pattern because the route boundaries are obvious.

​
Pattern 2: split by function

Examples:
  • one bot for support Q&A
  • one bot for build/deploy or ops workflows
  • one bot for CRM/calendar/tool-connected tasks
This can work well, but only if users know which one to invoke and where.

​
Pattern 3: split by environment

Examples:
  • one bot for prod/community
  • one bot for staging/testing
This is useful when you want safer experimentation without contaminating the real bot behavior.

​
Risky pattern: overlapping bots in the same channels with vague trigger rules

This is how you get:
  • duplicate responses
  • no responses
  • confusion about mentions
  • people assuming the wrong bot is broken
Avoid this if you can.

​
Fast troubleshooting flow

​
1) Map one working route per bot

For each bot or agent, document one known-good path:
  • server
  • channel
  • thread behavior
  • whether mention is required
  • what kind of requests it should answer
If you cannot produce that list, the setup is under-defined.

​
2) Compare permissions across the bots

A lot of “Bot B is broken” reports are really:
  • Bot A has read/reply access
  • Bot B is online but missing one key permission
Compare:
  • read access
  • send/reply access
  • thread visibility
  • mention/event handling
  • channel-specific restrictions
Online presence alone tells you almost nothing here.

​
3) Test bots one at a time in controlled contexts

Do not test a new bot by tossing it straight into the busiest shared channel. Instead:
  • test in a DM or clearly scoped test channel
  • send one explicit prompt
  • verify expected response behavior
  • then expand scope
That gives you signal instead of Discord chaos.

​
4) Avoid ambiguous trigger conditions

If two bots are in one place, be explicit about:
  • mention rules
  • response expectations
  • which one should stay quiet unless directly invoked
Otherwise people interpret silence as failure and overlap as intelligence. Neither interpretation is reliable.

​
5) Check gateway health, but do not stop there

Yes, run: 
openclaw status

openclaw gateway status --json
But if the gateway is healthy and only some bots/routes misbehave, you are probably debugging scope or routing, not a total runtime failure.

​
Common root causes

​
Root cause 1: unclear route ownership

Symptoms:
  • bots overlap
  • users do not know which bot should answer
  • failures seem random
Fix:
  • define channel/task ownership clearly

​
Root cause 2: permission differences between bots

Symptoms:
  • one bot works, another does not
  • both appear online
Fix:
  • compare permissions directly

​
Root cause 3: shared-channel ambiguity

Symptoms:
  • duplicate replies
  • missing replies
  • inconsistent behavior in busy channels
Fix:
  • tighten mention/trigger rules
  • reduce overlap

​
Root cause 4: staging/prod confusion

Symptoms:
  • the wrong bot is being tested
  • changes seem to affect the wrong environment
Fix:
  • separate staging and production paths more clearly

​
Root cause 5: route-specific config drift

Symptoms:
  • setup worked before adding another bot
  • one path regressed after a config change
Fix:
  • compare working and broken routes side by side

​
Practical checklist

Use this order:

​
1. Define ownership for each bot/agent

​
2. Identify one known-good route per bot

​
3. Compare permissions and scope

​
4. Test each bot in isolation first

​
5. Tighten trigger/mention rules in shared channels

​
6. Check openclaw gateway status --json

​
Final thought

Running multiple OpenClaw bots or agents in one Discord setup is absolutely doable. But the moment you add the second one, you are no longer just doing installation. You are doing traffic design. If that traffic design stays fuzzy, the setup will feel flaky even when the software is technically healthy.
  • Connected Channel or Bot Is Online but OpenClaw Ignores Messages
  • Healthy Gateway, Wrong Routing: Why OpenClaw Isn’t Responding in the Channel You Expect
  • Why OpenClaw Says It Will Act but Doesn’t Complete the Task