create_agentLABSCO SUMMARY
Specialists that persist, rather than a fresh persona each time.
Persistence is what separates this from writing a role into the prompt: an agent keeps its instructions and its history between sessions, so ask the risk analyst means the same thing tomorrow as it did today.
What it is
A Python MCP server for multi-agent orchestration on Claude. Agents are created with their own instructions, stored locally so they survive a restart, and addressed by name; one call builds a four-role financial analysis team.
What you get
- Agents created with a name and custom instructions, and deleted when finished with
- Every agent listed
- A message sent to a named agent, and its reply returned
- Conversation history kept per agent, and resettable
- A finance team created in one call: risk analyst, portfolio manager, data analyst, research analyst
- Agents and conversations stored locally, so they survive a restart
Requirements
Python 3.10+, Claude Desktop, and an Anthropic API key with billing enabled, supplied through the environment. Every agent call bills to that key.
Setup effort
Build from source — git clone https://github.com/Mayank1805/claude_swarm_mcp_agent
USE CASES
01
Setting up a specialist you will come back to
Create the agent with the instructions that define it, and confirm it is in the list.
02
Getting a second opinion from a different point of view
Put the same question to an agent with different instructions and compare the answers.
03
Standing up a finance team without configuring four agents
One call builds the risk, portfolio, data and research roles together.
04
Picking up where a specialist left off
Read that agent's conversation history, then continue it rather than re-explaining.
05
Starting a topic over without losing the agent
Reset the conversation and keep the instructions that make the agent useful.
06
Clearing out agents you no longer use
List what exists and delete the ones that were built for a finished piece of work.
7 tools
Listed in project docs. Your client may show a different list.
list_agentschat_with_agentdelete_agentcreate_finance_teamget_conversation_historyreset_conversationBEFORE YOU INSTALL
What to know first
Tools you get
7 tools named in the project's documentation.
Requires
- Python 3.10+
- ANTHROPIC_API_KEY in your environment
Where it runs
The documented client runs the server on your machine.
INSTALLATION
Connect your client
You clone this one yourself.
Server setup
Clone the repository
TERMINAL
git clone https://github.com/Mayank1805/claude_swarm_mcp_agent
cd claude_swarm_mcp_agentInstall dependencies
TERMINAL
pip install -r requirements.txtMCP Clients
Claude Desktop
CLAUDE DESKTOP · JSON
{
"mcpServers": {
"claude-swarm": {
"command": "python3",
"args": [
"<YOUR_PATH_TO_CLAUDE_SWARM_MCP>/claude_swarm_mcp_server.py"
],
"env": {
"ANTHROPIC_API_KEY": "<YOUR_ANTHROPIC_API_KEY>"
}
}
}
}
