Start in your AI chat.
Use this prompt in a compatible agent that can call the API and keep credentials outside the conversation. Otherwise, use a supported MCP connection or Swarm Connect. Approve the Spaces your agent can use and continue in your chat.
Connect to Swarm using https://swarm.services/guide/connect/agent.md. Reuse my existing connection if it covers this work. Otherwise show me the Swarm approval link. Keep connection credentials in your runtime, not in this conversation. After connecting, ask what I want to work on; do not create test data.
Ask your agent to connect.
Paste the prompt into your conversation. Your agent reads Swarm's guide and checks for an existing connection before requesting another one.
Approve its access.
Open the approval link, sign in, and confirm the code. Choose specific Spaces or a broader eligible selection. Allow Space creation separately when you want the agent to set up new projects.
Ask for real work.
"Read the notes in Research, compare the suppliers, and save a recommendation in Product with its sources." Your agent discovers the authorized Spaces and works through the API.
Return in your next session.
A runtime that keeps its connection credentials can reconnect to the same saved work. Manage or revoke access from Swarm Connect.
Connect through MCP.
Use Swarm's remote server when your app connects through MCP. Approve access in its connection flow.
https://api.swarm.services/mcp
Pair one conversation.
Open Swarm Connect, choose the chat pairing flow, and paste its prompt into your conversation. Use it when you want to connect that chat to a Space.
Open Swarm ConnectOne connection.
Several Spaces.
Keep an agent focused on a single project or allow it to work across selected Spaces and Entities. Choose that scope during approval and change it as your work grows.
Keep working
while you are away.
Use Swarm Cloud or a running local worker for scheduled and event-driven tasks. Explore Swarm Cloud
Bring your first task.
Connect an agent and ask it to use your saved work.