Connect AI Agents to Edge Delta with the MCP Server

Overview

The Edge Delta MCP (Model Context Protocol) server connects AI agents to your Edge Delta organization. Any MCP-capable client, such as Claude Code, Claude Desktop, Cursor, or VS Code, can use it to search logs, analyze log patterns, query metrics and traces, investigate events, inspect dashboards, and manage pipeline configurations through natural language.

Edge Delta hosts the server, and every API token comes with a personal remote MCP endpoint, so there is nothing to install: you create a token, copy the URL, and paste it into your MCP client. The server exposes 28 tools covering observability queries, pipeline management, live capture, and processor dry runs. See the Tools Reference for the complete list.

The following diagram shows how an MCP client reaches your Edge Delta data:

graph LR
  A[Claude Code] --> D[Hosted MCP server
api.edgedelta.com]
  B[Cursor] --> D
  C[Other MCP clients] --> D
  D --> F[Edge Delta API]
  F --> G[Logs, metrics, traces,
events, patterns]
  F --> H[Pipelines and dashboards]

Note: This page covers giving external AI clients access to Edge Delta. For the reverse direction, where Edge Delta’s own AI teammates access your telemetry data, see the Edge Delta MCP connector. For MCP as a concept, see the Model Context Protocol guide.

Get your MCP endpoint

Create an API token to get your remote MCP endpoint:

1. In the Edge Delta web app, click Admin > My Organization and open the API Tokens tab.
2. Click Create Token and give the token a name.
3. Pick a permission preset. Read & Write is recommended for MCP use: the agent can query everything and create or update resources such as pipelines, monitors, and dashboards, with no organization administration access. Choose Read Only for agents that should only investigate, or Custom to hand-pick permissions.
4. Click Create. The token is shown once, together with a Use this token section.
5. Open the MCP server tab and copy the remote MCP endpoint.

The URL embeds your organization ID and the token, so it is ready to paste into any MCP client:

https://api.edgedelta.com/v1/orgs/YOUR_ORG_ID/mcp?token=YOUR_API_TOKEN

Warning: The URL contains your API token. Treat the full URL as a secret: do not commit it to version control or share it in chat, and revoke the token under Admin > My Organization > API Tokens if it leaks.

Connect your AI client

Use the URL you copied in the client of your choice.

Claude Code

Register the server with one command:

claude mcp add --transport http edgedelta \
  "https://api.edgedelta.com/v1/orgs/YOUR_ORG_ID/mcp?token=YOUR_API_TOKEN"

Claude Desktop and claude.ai

Add a custom connector under Settings > Connectors > Add custom connector and paste the URL.

Cursor

Add the server to ~/.cursor/mcp.json (or .cursor/mcp.json in a project). This is the same configuration the Create Token dialog generates for you:

{
  "mcpServers": {
    "edgedelta": {
      "url": "https://api.edgedelta.com/v1/orgs/YOUR_ORG_ID/mcp?token=YOUR_API_TOKEN"
    }
  }
}

VS Code

Add the server to your mcp.json:

{
  "servers": {
    "edgedelta": {
      "type": "http",
      "url": "https://api.edgedelta.com/v1/orgs/YOUR_ORG_ID/mcp?token=YOUR_API_TOKEN"
    }
  }
}

Other MCP clients

Any client that supports streamable HTTP transport can use the URL directly. If your client supports custom headers and you prefer not to put the token in the URL, connect to https://api.edgedelta.com/v1/orgs/YOUR_ORG_ID/mcp and send the token in an X-ED-API-Token header instead.

Verify the connection

After configuring your client, confirm the integration works:

1. Restart the client so it picks up the new server.
2. Check that the client lists the edgedelta server and its tools. In Claude Code, run /mcp.
3. Ask a test question, for example: “What log patterns have negative sentiment in the last hour?”

You can also inspect the server outside a client with the MCP Inspector by entering your endpoint URL as a streamable HTTP server:

npx @modelcontextprotocol/inspector

OAuth endpoint

Clients that support MCP OAuth can connect to https://api.edgedelta.com/mcp instead of the tokenized URL. The client discovers the Edge Delta authorization server automatically and opens a browser login; access follows your Edge Delta account and no long-lived token is stored. For example, in Claude Code:

claude mcp add --transport http edgedelta https://api.edgedelta.com/mcp

Self-host with Docker

You can also run the server yourself instead of using the hosted endpoint, for example to pin a version, to keep traffic inside your own environment, or for clients that only support stdio servers. The image is published at ghcr.io/edgedelta/edgedelta-mcp-server:latest, and the source is available at github.com/edgedelta/edgedelta-mcp-server.

The container communicates over stdio, which is what desktop MCP clients expect, and reads its configuration from environment variables:

Configure a stdio client with the Docker command. For example, in Claude Desktop’s claude_desktop_config.json:

{
  "mcpServers": {
    "edgedelta": {
      "command": "docker",
      "args": [
        "run", "-i", "--rm",
        "-e", "ED_ORG_ID",
        "-e", "ED_API_TOKEN",
        "ghcr.io/edgedelta/edgedelta-mcp-server:latest"
      ],
      "env": {
        "ED_API_TOKEN": "YOUR_API_TOKEN",
        "ED_ORG_ID": "YOUR_ORG_ID"
      }
    }
  }
}

The same command works in any stdio-capable client. The binary also supports a streamable HTTP mode (... edgedelta-mcp-server:latest http, port 8080 by default, ED_MCP_PORT overrides) for shared deployments.

Best practices

Apply the usual credential hygiene to MCP configurations:

• Use a dedicated API token per MCP client so you can rotate or revoke each one independently.
• Scope tokens to what the agent needs. The Read Only preset prevents accidental pipeline changes; with Read & Write, the deploy_pipeline tool can change what your fleet runs.
• The endpoint URL and client configuration files contain the token in plain text; keep project-level MCP config files out of version control.

Troubleshooting

See Also

• Edge Delta MCP Server Tools Reference - Every tool with parameters and example prompts.
• Edge Delta MCP connector - Give Edge Delta AI teammates access to your telemetry data.
• edx CLI - Terminal access to the same platform surface, including AI agent skills.
• Create an API Token - Token management details.

In this section