Use d5s from Claude, ChatGPT, or Cursor

Connect the d5s MCP server to Claude, ChatGPT, Cursor, or another tool that supports remote MCP servers. Your tool can then use your d5s workspaces for you. It can start chats with your agents, read results, answer held actions, read and add files, run automations, and refresh dashboards.

Before you start

• You need a d5s account. The tool acts as you and sees only what you can see in d5s.
• Use the server URL https://api.eu.d5s.tech/mcp.
• You sign in to d5s once, in your browser, when you add the server. You do not need an API key.

Connect your tool

Claude and Claude Desktop

1. Open Settings → Connectors.
2. Select Add custom connector.
3. Enter the name d5s and the server URL.
4. Select Connect and sign in to d5s.

Claude Code

Run this command, then run /mcp in Claude Code and sign in:

claude mcp add --transport http d5s https://api.eu.d5s.tech/mcp

ChatGPT

1. Turn on developer mode in Settings → Apps → Advanced settings.
2. Select Create app and enter the server URL.
3. Select OAuth as the authentication and sign in to d5s.

Cursor

Add the server to ~/.cursor/mcp.json, then sign in when Cursor asks:

{
  "mcpServers": {
    "d5s": { "url": "https://api.eu.d5s.tech/mcp" }
  }
}

What your tool can do

Area	Tools
Account	identity, list_workspaces
Workspace	list_agents, list_projects, get_project, search
Chats	list_conversations, create_conversation, create_conversation_message, get_conversation_messages, stop_run
Approvals	permissions_list_open, permissions_respond
Files	files_list, files_cat, files_grep, files_create
Automations	list_automations, run_automation, list_automation_runs
Dashboards	list_dashboards, get_dashboard, refresh_dashboard

Each call names a workspace. Ask your tool to call list_workspaces first.

Each d5s tool tells your AI tool whether it only reads or also makes changes. Your AI tool can use this to ask your approval only for changes.

Each agent has its own chat. To read it, pass the agent ID from list_agents to list_conversations.

Follow a run

Sending a message starts an agent run and returns at once. A run can take many minutes. Your tool reads get_conversation_messages to see the run status:

• running: the agent is still working.
• needs_you: the agent is waiting for a decision. It can wait for an action approval, a connector tool approval, or an answer to a question. Use permissions_list_open and permissions_respond for approvals. Post a message to answer a question. Without an agent ID, your message goes to the agent that wrote the last message.
• done, failed, or stopped: the run ended.

stop_run stops a running agent. run_automation and refresh_dashboard also start runs that you follow the same way.

Limits

• Lists return 25 items for each page. Use offset to read the next page.
• files_cat returns up to 500 lines or 20 KB of text, and images up to 2 MB.
• files_cat and files_grep read text files up to 10 MB.
• files_grep returns up to 50 matching lines.
• files_create adds a text file of up to 50 KB to a chat.
• search returns 10 results by default and up to 50.
• get_dashboard returns up to 500 rows for each dataset.

Troubleshooting

• The tool asks you to sign in again. Your sign-in expired. Sign in again from the tool.
• A call returns d5s API error 403 or 404. You do not have access to that workspace or item in d5s. Check the workspace ID from list_workspaces.
• A call returns d5s API error 402. Your organization has no credits left for new runs. See Plans and credits.
• You cannot connect. Check that you use the server URL https://api.eu.d5s.tech/mcp.