Busabase x Knowledge Base for Coze SpaceKnowledge Base for Coze Space
Coze Space can split a complex agent into connected nodes with independent prompts, tools, workflows, knowledge, and routing conditions. That makes configuration legible, but a routed answer is still an execution result rather than an approved organizational fact.
MAP
Routing decides who handles a task; review decides what the organization should remember.
Coze Space can split a complex agent into connected nodes with independent prompts, tools, workflows, knowledge, and routing conditions. That makes configuration legible, but a routed answer is still an execution result rather than an approved organizational fact.
Persona, variables, triggers, databases, shortcuts
Prompt, use case, model, tools, knowledge
Start node, previous responder, global jump
A tested response or artifact
BOUNDARY
Global settings and node settings have different inheritance scopes
Routing decides who handles a task; review decides what the organization should remember.
| Global configuration | Persona, variables, triggers, databases, shortcuts | Inherited by the agent graph |
|---|---|---|
| Agent node | Prompt, use case, model, tools, knowledge | A scoped execution role |
| Routing condition | Start node, previous responder, global jump | A dispatch rule with precedence |
| Published outcome | A tested response or artifact | Evidence for review, not canonical truth |
SCOPE
Agent nodes can carry independent plugins, workflows, and knowledge
Coze Space can split a complex agent into connected nodes with independent prompts, tools, workflows, knowledge, and routing conditions. That makes configuration legible, but a routed answer is still an execution result rather than an approved organizational fact.
Inherited by the agent graph
A scoped execution role
A dispatch rule with precedence
Evidence for review, not canonical truth
EVIDENCE
Global jump conditions have priority over node applicability
Global settings and node settings have different inheritance scopes Agent nodes can carry independent plugins, workflows, and knowledge Global jump conditions have priority over node applicability
| Global configuration | Persona, variables, triggers, databases, shortcuts | Inherited by the agent graph |
|---|---|---|
| Agent node | Prompt, use case, model, tools, knowledge | A scoped execution role |
| Routing condition | Start node, previous responder, global jump | A dispatch rule with precedence |
| Published outcome | A tested response or artifact | Evidence for review, not canonical truth |
ACCEPTANCE
Routing decides who handles a task; review decides what the organization should remember.
Global settings and node settings have different inheritance scopes Agent nodes can carry independent plugins, workflows, and knowledge Global jump conditions have priority over node applicability
Verify Persona, variables, triggers, databases, shortcuts before accepting Inherited by the agent graph.
Verify Prompt, use case, model, tools, knowledge before accepting A scoped execution role.
Verify Start node, previous responder, global jump before accepting A dispatch rule with precedence.
Verify A tested response or artifact before accepting Evidence for review, not canonical truth.
FAQ
Knowledge Base for Coze Space: practical questions
Routing decides who handles a task; review decides what the organization should remember.
Does a multi-agent route prove the answer is correct?
No. It proves which node handled the input. The answer still needs evidence and a reviewer.
Are tools and knowledge attached to every node?
No. They can be configured per Agent node, so the active node and version matter.
What should be recorded from a published Agent?
Record the agent type, configuration version, source evidence, and tested behavior, not only its final prose.
Can a copied node be treated as an independent policy?
Only after its prompt, tools, knowledge, and routing context are reviewed as a new scope.
NEXT
Routing decides who handles a task; review decides what the organization should remember.
Coze Space can split a complex agent into connected nodes with independent prompts, tools, workflows, knowledge, and routing conditions. That makes configuration legible, but a routed answer is still an execution result rather than an approved organizational fact.