Ingest new customer support tickets
Pull latest tickets from the support queue for the triage agent to classify by urgency and topic.
AI Agents
Multi-Agent Orchestration Board
Coordinate AI Agent Handoffs with a Shared Kanban Board
Multi-agent systems fail most often at the handoff, not the task itself. This Multi-Agent Orchestration Board template gives every agent in your pipeline a shared source of truth: a kanban board where tasks move through Backlog, Assigned, In Progress, Handoff Review, and Done. Agents use Kanboard's MCP server to list, claim, update, and move tasks, logging the context each next agent needs directly in task descriptions. This template is built for teams running research agents, support triage agents, data enrichment agents, or any pipeline where work passes between multiple autonomous or semi-autonomous agents and needs a clear audit trail.
kanboard.io/templates/multi-agent-orchestration
Template preview
Multi-Agent Orchestration Board
5 columns
Backlog
3Holds tasks not yet assigned to any agent, awaiting triage or dependency resolution.
Draft weekly competitor summary
Research agent to gather competitor pricing and feature changes from the past 7 days.
Validate scraped product data
Data-validation agent checks scraped catalog entries for missing fields before enrichment.
Assigned
3Tasks matched to a specific agent role, waiting for that agent to pick them up.
Classify support tickets by urgency
Assigned to triage agent; output should tag each ticket with priority and route to correct queue.
Generate outreach email drafts
Assigned to copywriting agent; drafts should reference lead's last interaction from CRM notes.
Cross-check invoice totals
Assigned to finance-audit agent; flag any invoice with a mismatch over $50.
In Progress
2Tasks an agent has started working on, tracked here until output is ready for review.
Summarize research findings on market sizing
Research agent compiling findings into a structured brief with sources cited.
Enrich product catalog with missing attributes
Enrichment agent filling in size, material, and category fields using vendor API data.
Handoff Review
3Completed agent outputs awaiting validation or context handoff before being marked done.
Review generated outreach emails for tone
Reviewer agent checks drafts against brand voice guidelines before sending to sales team.
Verify ticket classifications before routing
QA agent spot-checks triage agent's urgency tags for accuracy.
Confirm invoice discrepancy report is complete
Finance lead reviews flagged invoices and confirms next steps documented.
Done
2Fully completed tasks with all handoffs closed, kept for audit trail and agent performance review.
Weekly competitor summary delivered
Final report shared with product team; archived with sources and agent notes.
Support ticket batch fully triaged and routed
All tickets from this batch classified, routed, and confirmed received by support leads.
How to run this board
Step 1
Route new work into Backlog with clear task descriptions and any known dependencies so agents have enough context to start.
Step 2
Assign tasks to specific agent roles by moving them into Assigned, tagging which agent or agent type should pick up the work.
Step 3
Agents move their claimed tasks to In Progress and use update_task to log intermediate context, decisions, and blockers as they work.
Step 4
Completed outputs move to Handoff Review where a reviewer agent or human validates quality and confirms handoff notes are complete.
Step 5
Once validated, tasks move to Done, preserving the full context trail for auditing agent performance and debugging failed handoffs.
AI agent usage
Run it with AI agents
This playbook helps AI agent teams use Kanboard's MCP server to self-organize multi-agent work: agents pull tasks from a shared board, respect column-based state transitions, and log handoff context in task descriptions so downstream agents never lose context between steps.
# Multi-Agent Orchestration Board
Columns: Backlog, Assigned, In Progress, Handoff Review, Done.
Rules for agents:
1. Use list_tasks to find tasks in 'Assigned' matching your agent role.
2. Call move_task to shift a task to 'In Progress' before starting work.
3. Append your output and any context the next agent needs to the task description via update_task before moving a task to 'Handoff Review'.
4. A human or reviewer agent moves tasks from 'Handoff Review' to 'Done'.
5. Never delete tasks unless explicitly instructed; use delete_task only for duplicates.List all tasks currently in the 'Handoff Review' column and summarize what context each one has for the next agent.
Move task 'Summarize research findings' from Assigned to In Progress and add a note about which agent is working on it.
Create a new task in Backlog for the data-cleaning agent, including the dataset name and expected output format.
Learn more about the board-centric workflow in Kanban for AI Agents or open the MCP guide.
Frequently asked questions
How many agents or tasks can I manage on the free plan?
The free plan supports 1 project and up to 25 tasks total, which is enough to run a small multi-agent pipeline. If your orchestration grows beyond that, upgrading to pro ($12/mo or $96/yr) removes those limits.
Can agents update tasks directly without a human in the loop?
Yes. Agents can use the MCP server tools like update_task and move_task to log progress and shift tasks between columns autonomously, though we recommend keeping a Handoff Review column for validation.
Does this template require a specific number of columns?
This template ships with five columns, but you can adjust between 3 and 6 columns depending on how many handoff stages your agent pipeline needs.
How do agents know which task belongs to them?
Use list_tasks to query tasks in the Assigned column and check the task description or title for the agent role it was tagged for, then move it to In Progress once claimed.
Can I access this board outside of the API?
Yes, you can view and manage the same board through the kanban board interface, keyboard shortcuts, command palette, or the macOS app, in addition to the REST API and MCP server.
What happens if a task exceeds the 25-task free limit?
You'll need to complete or archive existing tasks to stay within the free plan's 25-task cap, or upgrade to pro for unlimited tasks across your projects.
Related templates
Ready to create this board?
Sign in and use the template to create a project with columns and sample tasks.