Hey there, your guide is ready.
7 AI AGENT
TYPES IN
CLAUDE MANAGED AGENTS

Launched April 2026. No servers. No infrastructure. No setup beyond an API key. Here are all 7 types — what each does and exactly how to build them.

7Agent Types
90sFirst Build
£0Infra Cost
⬇️  Download as PDF
Free. Keep it forever. No catch.

What is Claude Managed Agents?

Anthropic's hosted platform for running autonomous AI agents — launched April 8, 2026. You define the agent (model + system prompt + tools). Anthropic runs it in a secure cloud container. No servers to build, no infrastructure to maintain. All 7 types below run on it.

⚡ One-time setup — works for all 7 types
1
Get your free API key
platform.claude.com/settings/keys
2
Install the Python SDK
pip install anthropic
3
Set your API key
export ANTHROPIC_API_KEY="your-key-here"
4
The SDK sets the required beta header automatically
managed-agents-2026-04-01 ← handled for you
01
🔧
Basic Agent With Tools
Beginner Friendly

Create an agent with the built-in toolset, start a session, give it a task. Anthropic spins up a secure cloud container and the agent runs bash commands, searches the web and reads/writes files — all autonomously on Anthropic's infrastructure.

Think of it as
"A personal assistant with a terminal, a browser and your files — running in the cloud 24/7."
Use Cases
Run scripts and shell commands autonomously in a managed container
Search the web, fetch pages, read and write files without watching it
Any long-running task needing a terminal, a browser and a file system
How to build it — Claude Managed Agents
# 1. Create your agent once — save the returned agent.id
ant beta:agents create \
  --name 'My Assistant' \
  --model '{id: claude-opus-4-7}' \
  --system 'You are a helpful assistant.' \
  --tool '{type: agent_toolset_20260401}'
# 2. Create a cloud environment — save the returned environment.id
ant beta:environments create \
  --name 'my-env' \
  --config '{type: cloud, networking: {type: unrestricted}}'
# 3. Start a session and send your task — streams back in real time
session = client.beta.sessions.create(agent=agent_id, environment_id=env_id)
client.beta.sessions.events.send(session.id,
    events=[{type: 'user.message', content: 'Your task here'}])
02
🔌
MCP Server Agent
⚡ This is where it gets powerful

Add MCP (Model Context Protocol) server URLs when you create your agent. It gains live read/write access to those external services — Notion, Atlassian, Slack, any database with a public MCP URL — and can interact with everything stored there.

Think of it as
"Giving your agent a keycard to every system in your business, not just its own container."
Use Cases
Connect to Notion and update pages or databases automatically
Read Jira boards, create tickets and update statuses hands-free
Plug into any service that exposes a public MCP server URL
How to build it — Claude Managed Agents
# Same as Type 01 but add --mcp-server when creating the agent
ant beta:agents create \
  --name 'Notion Agent' \
  --model '{id: claude-opus-4-7}' \
  --system 'You read and write Notion pages.' \
  --tool '{type: agent_toolset_20260401}' \
  --mcp-server '{type: url, url: https://mcp.notion.com/sse}'
# Then start a session exactly as in Type 01.
# Add multiple --mcp-server flags for multiple services.
03
🏭
Sequential Agents
AI Assembly Line

Create separate agents for each stage of your pipeline. Start sessions in order, passing the output of each as the input to the next. Each agent is a specialist that only does its one job, perfectly, before handing off.

Think of it as
"A factory production line — each station does one specialist job, then hands off to the next."
Use Cases
Research → Summarise → Write → Publish, each step a different agent
Prospect → Qualify → Draft outreach → Send, fully automated end-to-end
Ingest raw data → Clean → Analyse → Generate report
How to build it — Claude Managed Agents
# Create three specialist agents, each with its own system prompt
agent_a = create_agent('You read and summarise contacts.')
agent_b = create_agent('You write personalised emails.')
agent_c = create_agent('You send emails via the API.')
# Run in sequence — output of each becomes input of the next
session_a = start_session(agent_a)
result_a = send_and_wait(session_a, original_task)
session_b = start_session(agent_b)
result_b = send_and_wait(session_b, result_a)
session_c = start_session(agent_c)
final = send_and_wait(session_c, result_b)
04
Parallel Execution Agents
Speed Multiplier

Start multiple sessions at the same time — each runs independently in Anthropic's cloud. Use Python asyncio to fire them all at once and gather results when they're done. Massively faster for tasks you can split into parallel workstreams.

Think of it as
"Running three researchers at once instead of waiting for one to finish before starting the next."
Use Cases
Research three competitors simultaneously, merge findings into one report
Write five email variants at once, pick the best performer
Analyse separate datasets in parallel and combine the insights
How to build it — Claude Managed Agents
import asyncio
# Async helper — starts a session, waits for the result
async def run_session(task):
    session = client.beta.sessions.create(agent=agent_id, environment_id=env_id)
    return await send_and_wait_async(session, task)
# Fire all sessions at the same time — Anthropic runs them in parallel
results = await asyncio.gather(
    run_session('Research competitor A'),
    run_session('Research competitor B'),
    run_session('Research competitor C'))
merged = combine(results)
05
🚦
Agents With Routers
AI Traffic Controller

Create a lightweight router agent whose only job is to classify an incoming task and return a routing decision. Your code reads that decision and starts the right specialist session. Fully automatic, zero manual sorting.

Think of it as
"An air traffic controller who instantly decides which runway every inbound flight goes to."
Use Cases
Urgent tasks go to a priority queue, normal tasks to standard flow
Support tickets classified by complexity, routed to the right specialist
Leads sorted and sent to different sales workflows automatically
How to build it — Claude Managed Agents
# Router: classify input, return one word decision
router = create_agent(system='Read the task. Reply with ONE word: PRIORITY, STANDARD or ESCALATE.')
# Specialist agents for each route
priority_agent = create_agent('Handle urgent tasks fast.')
standard_agent = create_agent('Handle normal tasks carefully.')
escalate_agent = create_agent('Handle complex edge cases.')
# Route automatically — start the right specialist session
route = send_and_wait(start_session(router), incoming_task)
agents = {'PRIORITY':priority_agent,'STANDARD':standard_agent,'ESCALATE':escalate_agent}
result = send_and_wait(start_session(agents[route]), incoming_task)
06
🛡
Human in the Loop
⚠️ High-Stakes Tasks

Set your agent's tools to always_ask confirmation mode. Before the agent executes any action, your code receives a confirmation event. You approve or deny — the action only fires if you say allow. Full control, zero surprise executions.

Think of it as
"A surgeon's assistant who preps everything perfectly, lays out every tool, then waits for the surgeon to say go."
Use Cases
Sending large payments or invoices after you review the agent's draft
Mass email sends where a human must approve before it fires
Deleting, archiving or modifying critical data — only with your say-so
How to build it — Claude Managed Agents
# Create agent with always_ask tool confirmation
agent = create_agent(tools=[{type:'agent_toolset_20260401', confirmation:'always_ask'}])
# Stream events — intercept before any tool fires
for event in stream:
    if event.type == 'agent.tool_use':
        print(f'Agent wants to: {event.name}')
        ok = input('Allow? y/n: ')
        client.beta.sessions.events.send(session.id, events=[{
            type:'user.tool_confirmation', tool_use_id:event.id,
            result:'allow' if ok=='y' else 'deny'}])
07
👑
Orchestrator Agent
★ Most Powerful — Research Preview
🔬
Research Preview. Request access free at claude.com/form/claude-managed-agents — Anthropic is onboarding teams now.

Create specialist sub-agents, then create an orchestrator that lists them as callable_agents. The orchestrator dynamically spawns threads to delegate work at runtime — it figures out who it needs on demand. You don't pre-configure the flow.

Think of it as
"A CEO who instantly decides to bring in a lawyer, accountant and designer for a problem — and does it on the spot, without being asked."
Use Cases
Complex tasks where sub-tasks are unknown until mid-execution
Business analysis requiring a research agent, legal agent and comms agent in parallel
Anything too complex to pre-plan — the system self-organises at runtime
How to build it — Claude Managed Agents
# 1. Create your specialist sub-agents
researcher = create_agent('You research topics thoroughly.')
writer = create_agent('You write clear, concise summaries.')
# 2. Create orchestrator — list sub-agents it's allowed to call
orchestrator = client.beta.agents.create(
    name='Engineering Lead', model='claude-opus-4-7',
    system='Delegate research and writing to your specialist agents.',
    tools=[{type:'agent_toolset_20260401'}],
    callable_agents=[{type:'agent',id:researcher.id},{type:'agent',id:writer.id}])
# 3. Start ONE session — it spawns sub-agent threads automatically
session = start_session(orchestrator)
# Threads run in parallel, orchestrator merges and returns results

Quick Comparison

#TypeBest ForLevel
01Basic Agent With ToolsBash, web search, file opsEasy
02MCP Server AgentExternal databases via MCP URLsEasy
03Sequential AgentsOrdered multi-step pipelinesMedium
04Parallel ExecutionSpeed-critical split workstreamsMedium
05Router AgentHigh-volume task sortingMedium
06Human in the LoopHigh-stakes tasks needing approvalMedium
07Orchestrator AgentSelf-organising multi-agent tasksResearch Preview