Skip to main content

Dcycle MCP Server

The Dcycle MCP (Model Context Protocol) Server lets AI assistants query and manage your operational data directly — organizations, facilities, invoices, vehicles, employees, workforce, purchases, travel, logistics, emissions, projects, decarbonization plans, LCA, custom KPIs, and more. With 102 tools (78 read + 24 write), 6 reference resources, and 3 guided workflow prompts, you can analyze environmental impact, track reduction progress, create operational records, and generate compliance reports through natural language.

What is MCP?

MCP is an open standard created by Anthropic that lets AI assistants securely connect to external data sources. Think of it like connecting Google Calendar to Claude — but for your operational data.

Connection Modes

The Dcycle MCP server supports two modes:

Remote (Recommended)

Connect directly to https://mcp.dcycle.io — no installation needed. Authenticate with your Dcycle account via OAuth 2.0.

Local (stdio)

Run the server on your own machine with a Dcycle API key. Useful for automation or custom deployments.

What Can You Do?

Organizations

Hierarchy, metrics, and group structure

Facilities

Offices, factories, warehouses with details

Invoices

Electricity, gas, water invoices with co2e

Vehicles

Fleet data — fuel, ownership, emissions

Employees

Employees and commuting data

Workforce

ESG social data — demographics, contracts, locations

Purchases

Scope 3 purchased goods & services

Business Travel

Travel records by transport type

Hotel Stays

Accommodation stays with emissions

Transports

Upstream/downstream transport records

Logistics

Shipping, hubs, clients, and reports

Wastes

Waste records by facility

Waste Water

Treatment records, discharge, co2e

Sold Products

Products, periods, per-country emissions

Files

Uploaded files and processing status

Emissions

GHG emissions, historical trends, emission factors

Projects

ESG projects, tasks, and activity feeds

Dashboards

Saved report dashboards with widgets

Decarbonization

Plans, targets, actions, and baselines

Custom KPIs

KPI datasets, campaigns, and values

LCA

Life Cycle Assessment portfolios and impacts

Elastic Data

Custom datasets with flexible schemas

Compliance

Environmental legislation and regulations

Reference Data

GHG scopes, categories, sectors

Available Tools (93)

Most tools are read-only. Write tools (create_facility, update_facility, create_purchase, update_purchase, create_business_travel, update_business_travel, create_hotel_stay, update_hotel_stay, create_transport, update_transport, create_waste, update_waste, create_decarb_target, create_decarb_action, create_project, create_project_task, submit_kpi_response) are annotated with readOnlyHint: false so AI assistants can distinguish them and confirm with you before making changes.

Resources (6)

In addition to tools, the server exposes MCP Resources — static reference data automatically available to the AI assistant as context. See Reference Resources for full details and content.

Prompts (3)

The server also provides MCP Prompts — guided workflows that chain multiple tools together to produce structured reports. Each prompt accepts a year argument (defaults to the previous calendar year).
Prompts are activated by the AI assistant when you ask for a report — e.g. “Generate a carbon footprint summary for 2025”. You don’t call them directly.

How It Works

Remote mode

Local mode (stdio)

Security

  • Authentication required — You must authenticate before any data is accessible. Remote connections use OAuth 2.0 (your existing Dcycle account login); local connections require a scoped API key.
  • Your data stays yours — All requests are scoped to organizations you already have access to in Dcycle. The MCP cannot access any org you don’t have permission for.
  • No credentials stored in the AI tool — In remote mode, your Dcycle password is entered on app.dcycle.io, not transmitted to the AI assistant. The AI receives only a short-lived Bearer token.
  • Industry-standard protocols — OAuth 2.0 with PKCE for browser flows, JWT validation via RS256, and HTTPS only for the remote server (mcp.dcycle.io).
  • Safe by default — Read tools are annotated with readOnlyHint: true. Write tools (create_facility, update_facility, create_purchase, update_purchase, create_business_travel, update_business_travel, create_hotel_stay, update_hotel_stay, create_transport, update_transport, create_waste, update_waste, create_decarb_target, create_decarb_action, create_project, create_project_task, submit_kpi_response) are annotated with readOnlyHint: false and destructiveHint: false, so AI assistants confirm before making changes. No delete operations are exposed.

Compatible AI Assistants

The Dcycle MCP server works with any MCP-compatible AI assistant. Here’s how to connect the most common ones:
Ask your Claude admin to add a custom connector:
  • Name: Dcycle
  • URL: https://mcp.dcycle.io/mcp
On first use Claude will open a browser for OAuth authentication.For local mode with an API key, see the Quickstart.
On first connection Claude Code will open a browser for OAuth authentication.For local mode: claude mcp add dcycle -- uvx dcycle-mcp (requires DCYCLE_API_KEY env var).
Requires VS Code 1.101+ with the GitHub Copilot extension. MCP tools are available in Agent mode only.Add to .vscode/mcp.json in your project (or open MCP: Open User Configuration from the command palette for a global config):
On first use, VS Code will open a browser window to authenticate with your Dcycle account via OAuth 2.0. Tokens are managed automatically by VS Code.
Note: The root key is "servers" — not "mcpServers" (which is the Claude Desktop format).
Requires a Pro, Plus, Business, or Enterprise plan. Configuration is done entirely through the UI — no config file needed.
  1. Go to Settings → Apps → Advanced settings → Developer mode and enable it.
  2. In any chat, click ”+” and select Add connector.
  3. Enter:
    • Name: Dcycle
    • MCP Server URL: https://mcp.dcycle.io/mcp
  4. On first use ChatGPT will redirect you to log in with your Dcycle account via OAuth 2.0.
Microsoft Copilot Studio is an enterprise no-code agent builder (separate from GitHub Copilot). MCP is supported in GA and requires a Copilot Studio license with Generative Orchestration enabled on your agent.
  1. Open your agent and go to the Tools tab.
  2. Click Add a tool → New tool → Model Context Protocol.
  3. Fill in the wizard:
    • Server name: Dcycle
    • Server description: Query Dcycle operational data — organizations, invoices, vehicles, employees, purchases, business travel, and logistics.
    • Server URL: https://mcp.dcycle.io/mcp
  4. For Authentication, select OAuth 2.0 → Dynamic discovery. Copilot Studio auto-discovers the Dcycle authorization endpoints.
  5. Click Create. Copilot Studio will show a Redirect URL — copy it.
  6. Register that Redirect URL in your Dcycle account settings so the OAuth flow can complete.
Copilot Studio discovers all available tools automatically at runtime. No manual tool registration is needed.
Note: Only Streamable HTTP transport is supported. Localhost servers are not reachable from Copilot Studio’s cloud infrastructure.
Add to ~/.gemini/settings.json (global) or .gemini/settings.json in your project root:
On first use Gemini CLI will open a browser for OAuth authentication. You can also trigger it manually:
Tokens are stored at ~/.gemini/mcp-oauth-tokens.json.
Note: Gemini in Google Workspace and Google AI Studio do not support connecting to external MCP servers. Use the Gemini CLI.
Cursor supports SSE transport only. Use mcp-remote as a bridge (requires Node.js):
~/.cursor/mcp.json
mcp-remote proxies the Streamable HTTP server to a local stdio process. OAuth is handled in the browser on first run.
Edit ~/.codeium/windsurf/mcp_config.json:
Windsurf supports OAuth 2.0 via automatic browser redirect on first connection.

Getting Started

Quickstart Guide

Connect in under a minute using the remote server — no installation needed

REST API

Build integrations with the REST API

CLI

Manage data from the command line