MCP
The OCPPLab API serves its operations as MCP tools. Agents such as Claude Code, Codex, Cursor, VS Code, claude.ai and ChatGPT can deploy virtual chargers, manage CPO and eMSP data, run OCPI roaming, and more, with your permissions.
- Transport: Streamable HTTP
- URL:
https://api.ocpplab.io/mcp, or one server per domain at/mcp/<domain> - Auth: browser sign-in (OAuth), or an SDK token as
Authorization: Bearer <token>
Connect in one command
Run /mcp in Claude Code and pick ocpplab. Your browser opens: sign in, choose the organization the agent may use, and select Allow. Other clients: Connect a client.
Pick the servers you need
Every connected server adds its tool list to the agent’s context. Connect only the domains your task needs:
Tool counts and every tool name, per server: Servers and tools.
How tools behave
- Names are
<domain>_<operation>, for examplecpo_charge_points_listorevcharger_start_session. - Permissions are yours: a tool call runs as the REST call it maps to, with your token. Results only include your organization, and admin tools return
403unless you are an admin. - Hints: every tool says whether it is read-only, destructive or idempotent, so your client can ask before a destructive call.
- Long-running tools return a tracking id, and their description names the tool to poll. See Long-running operations.
- Updates: omit a field to leave it unchanged, or send
nullto clear it. - Limits: 120 calls per minute per user and organization, shared by every MCP URL. Over it, tools return
429withRetry-After. - REST only: proxy backends, proxy routing rules and bulk log deletion are not MCP tools.
Agent skills
Give your coding agent OCPPLab know-how as agent skills: how to authenticate, deploy and drive virtual chargers, use the SDKs and the CLI, and connect MCP. One command installs them for Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, OpenCode and other agents:
Two MCP servers, two jobs
Connect both when you want an agent to look up how something works and then do it.