One agent with every tool makes a good demo; a team of agents with one job each is what holds up at work. Most multi-agent setups break at the hand-off: context is lost, two agents repeat the work, or one calls a tool it should never touch. Live, Param builds a planner, an engineer and a reviewer with the Claude Agent SDK, each scoped through an MCP gateway.
Why this matters
A single agent with every tool and one long prompt is the fastest way to a demo and the slowest way to something your team trusts. As the prompt grows it follows less of it, and one wrong tool call can touch anything. Splitting the work into roles, each with a narrow job and a short tool list, is how agent systems stay predictable as they grow. The hard part is the hand-off between them.
What happens in the hour
The problem4 min
Where multi-agent setups break: lost context at the hand-off, repeated work, a tool call that should never happen.
Built live40 min
Three roles with their own prompt and model, tools scoped through an MCP gateway, and one issue going from plan to pull request with every call logged.
Questions16 min
When a second agent pays for itself, compared with one agent on quality, latency and token cost.
What you will be able to do
Split one agent into roles
A planner, an engineer and a reviewer, each with its own prompt and model.
Scope tools per role
Each agent calls only the tools its role needs, and every call is logged.
Decide when it pays
Compare the team with one agent on quality, latency and token cost.
The ideas behind it
Planner, engineer, reviewer
Three roles that mirror how a team ships a change: one breaks the work down, one does it, one checks it before it lands.
Scoped tools through an MCP gateway
Each role sees only the tools it needs. The gateway enforces that, so a prompt cannot talk its way to a tool outside the role.
The hand-off
What one agent passes to the next: the task, the decisions already made and nothing else. Most multi-agent bugs live here.
When a second agent pays
More agents mean more calls and more latency. The lab compares one agent with three on quality, time and token cost on the same issue.
Where it fits
The parts of a production agent system, in the order the buildcamp builds them. The lit tiles are the ones this live lab builds; the Agentic AI Buildcamp for Engineers builds all of them.
Week 1Architect
Roles and modelsOne job per agent, a model chosen for that job, and a token budget.
OrchestrationA queue agents pick work from, with hand-offs a person can follow.
Context and memoryRetrieval with sources, memory across sessions, prompts laid out for the cache.
Tool callingTyped tools that act in GitHub, Linear and Google.
Week 2Build
MCP serversEvery tool behind an MCP gateway, scoped to the role that needs it.
Durable executionRuns that resume after a crash and never repeat a write.
Human in the loopA person approves anything that cannot be undone.
Week 3Secure and deploy
Identity and permissionsEach agent signs in as itself and acts on behalf of a user.
LLM gatewayRouting, caching and a budget on every model call.
Traces, evals and costEvery run traced, scored and charged to the agent that made it.
Multi-tenancyEach team or customer kept apart, in data and in the bill.
Week 4Govern and extend
Policy as codeRules checked on every action, not written in a document.
Signed skillsNew roles built from reviewed, signed skills.
Before you come
For
Engineers whose single agent has outgrown one prompt and one tool list, and architects deciding whether a multi-agent design is worth its cost.
You need
None. Code is shown, not required.
Track
Agentic AI for engineers who build and run it.
Questions
Which framework does the lab use?
The Claude Agent SDK, with tools served over MCP. The pattern of roles, scoped tools and explicit hand-offs carries over to LangGraph, CrewAI or your own loop.
Is multi-agent always better than one agent?
No. For short tasks one agent is cheaper and fast enough. The lab shows how to measure the difference so you can decide per workflow.
What is an MCP gateway?
A server that sits between your agents and their tools, serves the tools over the Model Context Protocol and decides which agent may call which tool.
Will I get the code?
Yes. Everyone who signs up gets the recording, and the code is shown in full during the build.
Is the Live Lab really free?
Yes. Live Labs are free on Maven. You sign up with your email and get the join link and the recording.
What if I cannot make it live?
Sign up anyway. Everyone who signs up gets the recording, so you can watch the build later and reply with questions.
Do I need to code along?
No. Most people watch the build and ask questions. Every step is shown, so you can repeat it on your own afterwards.