Blocks
Agent Orchestrator
Coordinate multiple agents working together
Orchestrates multiple AI agents to work together on complex tasks. Supports different coordination patterns including sequential, parallel, voting, and delegation.
Block Preview
Agent Orchestrator
Coordinate multiple agents working together
Usage
- Add the block to your workflow and connect it to the upstream step.
- Configure any required credentials or tokens in the inputs.
- Fill in required inputs and optional parameters for the run.
- Run a test execution, inspect outputs, and iterate before deploying.
- Deploy the agent_orchestrator block with monitoring enabled in production.
Inputs (UI)
Orchestration Pattern
dropdownLayout: half
Options: Sequential, Parallel, Voting, Delegation
Coordinator Model
dropdownLayout: half
Condition: orchestrationType = ["delegation","debate"]
Options: GPT-4o, GPT-4o Mini, Claude 3.5 Sonnet, Claude 3.5 Haiku
Coordinator Instructions
long-inputPlaceholder: Instructions for how the coordinator should manage the agents...
Layout: full
Condition: orchestrationType = ["delegation","debate"]
Task
long-inputPlaceholder: Describe the task for the agents to work on...
Layout: full
Agent Configuration
codeDefine the agents that will participate
Placeholder: [ { "name": "Researcher", "type": "specialist", "prompt": "You are a research specialist...", "model": "gpt-4o" }, { "name": "Writer", "type": "specialist", "prompt": "You are a writing specialist...", "model": "gpt-4o-mini" } ]
Layout: full
Shared Context
codeInitial shared memory available to all agents
Placeholder: { "initialData": {} }
Layout: full
Voting Threshold
sliderPercentage of votes needed for consensus
Layout: half
Condition: orchestrationType = "voting"
Max Debate Rounds
short-inputPlaceholder: 3
Layout: half
Condition: orchestrationType = "debate"
Timeout (seconds)
short-inputPlaceholder: 300
Layout: half
Enable Inter-Agent Messaging
switchAllow agents to communicate directly
Layout: half
Enable Shared Memory
switchAllow agents to read/write shared state
Layout: half
Result Aggregation
dropdownLayout: half
Options: Last Result, Merge All, Best (by coordinator), Consensus
On Agent Failure
dropdownLayout: half
Options: Continue, Abort All, Retry, Fallback
Inputs (API)
orchestrationType
stringRequired
coordinatorModel
stringOptional
coordinatorPrompt
stringOptional
task
stringRequired
agents
jsonRequired
sharedContext
jsonOptional
votingThreshold
numberOptional
maxRounds
numberOptional
timeout
numberOptional
enableMessaging
booleanOptional
enableSharedMemory
booleanOptional
aggregationStrategy
stringOptional
failureStrategy
stringOptional
Outputs
Primary response type:
{
"result": "json",
"agentResults": "json",
"messages": "json",
"consensus": "json",
"executionOrder": "json",
"totalDuration": "number"
}