←back to features

Multi-agent Orchestration

Coordinate multiple AI agents that share knowledge, delegate tasks, and work together on complex workflows. Build agent teams that are greater than the sum of their parts.

Real business problems - revenue forecasting, adversarial decision-making, continuous quality evaluation - demand multiple specialized agents working together. ChatBotKit handles multi-agent orchestration at the platform level, so you do not need a separate orchestration framework or custom coordination code.

Create multiple bots, assign them to shared Spaces for persistent knowledge, define their capabilities through Skillsets, and let the platform manage communication, delegation, and execution. Agents call other agents through skillset abilities, share context through Spaces, and coordinate through tasks and triggers.

Key Features

  • Shared Spaces: A persistent file layer where agents read what others have written and build on accumulated context.
  • Skillset-based messaging: Agents call other agents through skillset abilities, passing structured context and receiving results.
  • Task delegation: A supervisor agent assigns work to specialized sub-agents and merges their outputs.
  • Adversarial patterns: Deploy agents that argue for and against a proposal with a judge that synthesizes the result.
  • Asymmetric access: Different agents get different permissions on shared resources, enforcing governance boundaries.
  • Parallel execution: Run multiple agents simultaneously and merge results through a coordinator.
  • Scheduling and triggers: Control when each agent runs with configurable cadences and event-based triggers.

How It Works

All orchestration is managed through the platform. You create bots with different backstories and skillsets, connect them to shared Spaces, and configure tasks that trigger them on schedules or events. The platform handles message routing between agents, context passing, and execution ordering.

For example, a research agent can gather information and write to a shared Space. An analysis agent reads that output and produces insights. A reporting agent formats the result and posts it to Slack. Each agent has its own model, backstory, and permissions, but they all share the same knowledge layer.

For advanced use cases, it is also possible to build local multi-agent orchestration directly through the @chatbotkit/agent SDK, giving you full programmatic control over agent coordination, tool routing, and execution flow outside the platform.

The Agentic AI Blueprints feature page lists multi-agent blueprints you can fork and customize.

multi-agentsdkorchestrationagentsdevelopment