Overview

The Enterspeed Management MCP Server turns the Enterspeed Management API into a set of Model Context Protocol tools that AI clients (Claude, Cursor, VS Code, custom agents) can discover and call. It is the configuration counterpart to the Query MCP server: where Query MCP reads your data, Management MCP reads and changes your setup. You do not need to host anything yourself — Enterspeed runs the server for you. Typical uses:

Ask an agent to explain how a tenant is wired together — sources, environments, domains, clients, deployments.
Have an agent write a mapping schema, dry-run it against a real source entity, and save a new version.
Investigate a view or route that is not producing what you expect.
Query your tenant logs and metrics in natural language.

Hostname

Environment	URL	Transport
Production	`https://mcp.management.enterspeed.com/`	MCP Streamable HTTP

A health check is available without authentication:

curl -s "https://mcp.management.enterspeed.com/health"
# "ok"

The server speaks MCP Streamable HTTP on the root URL, and nothing else. Choose the HTTP or Streamable HTTP transport in your client. The older SSE transport does not work: a client set to SSE can complete the sign-in and then fail on its first request with an HTTP 400 error. There is no /sse endpoint — always configure the root URL.

How authentication works

You sign in with your existing Enterspeed account, over OAuth. The first time your client connects, a browser window opens and you authenticate with your normal Enterspeed credentials. The session then acts as you, with the permissions your account has in each tenant. An API key is refused as authentication on this server. Management API keys still exist, and this server has tools that list, rotate, and set an expiry on them, but a request that authenticates with an x-api-key header is turned away. Environment client keys belong to Query MCP.

Selecting a tenant

A sign-in names a person, not a tenant, so every signed-in session has to select a tenant before any tenant-scoped tool works — even if your account belongs to only one. The server’s set_tenant_context tool does this, and whoami lists the tenants your account belongs to. You rarely call either by hand; ask your client:

Which Enterspeed tenants do I have access to? Select the one called Acme.

The selection is held in memory by the server instance that handled it. If your client reconnects, or the server restarts, a tool call can come back saying no tenant is selected. Nothing is lost when that happens: select the tenant again and carry on. Your client lists the same tools before and after you select a tenant. A tenant-scoped tool called without a selection is still listed, and calling it fails against the Management API rather than telling you a tenant is missing, so run whoami if a call is refused and you are not sure which tenant the session is on. It reports the tenant in force, or tells you to select one. What the list depends on is the toolsets parameter (see Limiting a connection) and your account’s access level, never which tenant you picked.

What the server can and cannot do

The server exposes 40 tools, made up of 23 reads, 16 writes, and set_tenant_context, which selects the tenant for the session and changes nothing in it.

Reads cover the tenant overview, sources and source entities, mapping schemas and deployments, views and routes, indexes and index documents, domains and hostnames, destinations, environment and management clients, logs, and metrics. One read, diagnose_missing_content, combines several of them to find where content stops on its way to a URL.
Four writes are additive. create_tenant, create_mapping_schema, create_destination, and create_environment_client only ever add.
Twelve writes can change or replace something that already exists. They are the four upsert_* tools, update_environment_client, configure_destination, save_mapping_schema_version, deploy_mapping_schema, unpublish_mapping_schema, ingest_source_entities, rotate_management_client, and set_management_client_expiry. An upsert can change an existing resource as well as create one, configure_destination can change an environment’s existing destination setup, an unpublish stops an environment serving a schema’s views, a deploy changes what an environment serves, an ingest overwrites entities with the same origin id, and the two tools for management clients can end a key’s life early.

No tool deletes anything, by design. The server has no way to send a DELETE or a PATCH request at all. Every request it can make is a GET, or a POST/PUT on a fixed allow-list checked when the server starts. No tool can delete a tenant, schema, environment, source, source group, domain, destination, client, or entity.

In practice, an agent connected to this server cannot delete your configuration, but the twelve writes above can change what your tenant serves. If an agent only needs to look, open the connection read only. See Limiting a connection. See Tools for the full inventory with each tool marked read or write.

Limiting a connection

The server can only do what your account is allowed to do. To give an agent less than that, add parameters to the server URL when you configure your client. They are set by the person configuring the client, not by the agent, so nothing said in a conversation can lift them — you change them by reconnecting with a different URL.

Parameter	What it does	Example
`readOnly`	`true` makes every write tool refuse for the life of the connection. Reads, and selecting a tenant, still work.	`?readOnly=true`
`tenantId`	Pins the connection to one tenant, or to several separated by commas. Selecting any tenant outside the list is refused. Each value is a tenant’s id, in the form `gid://Tenant/{guid}`. `whoami` lists the ids of the tenants you belong to.	`?tenantId=gid://Tenant/xxxxxxxx-…` or `?tenantId=gid://Tenant/aaaaaaaa-…,gid://Tenant/bbbbbbbb-…`
`toolsets`	Offers only the named groups of tools. The groups are `core`, `schemas`, `sources`, `delivery`, `domains`, `destinations`, `clients`, `observability`, and `provisioning`, or `all`. `core` (who you are, tenant selection, and the tenant overview) is always included.	`?toolsets=schemas,delivery`

Combine them with &, and wrap the URL in double quotes on the command line so your shell does not split it:

"https://mcp.management.enterspeed.com/?readOnly=true&tenantId=gid://Tenant/xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx"

A value the server cannot read fails safe. Anything other than readOnly=true or readOnly=false, such as readOnly=1, an empty readOnly=, or the parameter given twice, makes the connection read only. A malformed tenantId refuses every tenant. So does a list with even one entry the server cannot read. The entries it can read are refused too, and the refusal names each entry it could not read. Give several tenants as one value with commas between them, because giving the parameter twice (?tenantId=…&tenantId=…) also refuses every tenant. An unknown group name in toolsets, or an empty value, offers core only. In each case the refusal names the parameter, and whoami reports how the connection was read. whoami reports the pinned tenants in connection.pinnedTenantIds. It is null when the connection is not pinned, the list of tenants when it is, and an empty list when the pin could not be read. Check it after connecting. The pin only works if your client sends the query string with every request, and a client that drops it opens an unpinned connection without saying so. Leave all three off and the connection offers every tool, on every tenant your account can reach. toolsets is not a security control — it decides which tools the client is offered, which also helps clients that cap how many tools they load across all servers.

Guardrails

Beyond your account’s own permissions and the connection parameters above, the hosted server applies three limits you should know about.

Every deploy and unpublish asks you to confirm first

deploy_mapping_schema and unpublish_mapping_schema ask you directly, in your client, before anything changes. The prompt names the schema, the versions and the environment, and you confirm by typing the environment’s name. Declining, cancelling, a name that doesn’t match, or no answer within about 50 seconds changes nothing.This applies to every environment, whatever it is called. You can deploy to any environment your account can deploy to in the Management App.The prompt needs a client that supports MCP elicitation, such as Claude Code, Codex, GitHub Copilot CLI, VS Code or Cursor. Clients that don’t, such as Claude.ai, Claude Desktop chat and Gemini CLI, are refused for these two tools; deploy from the Management App instead.

A client set to answer on your behalf will confirm without asking you. That includes auto-approve modes, Copilot’s Autopilot, and client hooks that answer prompts. Turn those off for connections that can deploy.

Access keys are never returned through MCP

The server strips the accessKey field out of every response before your client sees it, at every level of nesting. This covers environment client keys and source ingest keys.It applies even to the tools that mint a key. create_environment_client creates the client and rotate_management_client replaces a management client’s key, but the key value is in neither response. The agent is told a key was created and that it must be collected from the app — so it reports the right next step instead of inventing a value. Fetch the key in the Management App when you need it.The field is removed rather than masked, so no tool can observe a key at any point.

Rate limit — 60 requests per minute per person

Requests are counted per signed-in person, over a fixed one-minute window. Exceeding the limit returns:

HTTP 429 Too Many Requests
Retry-After: <seconds>

Too many requests. Retry after <n> seconds.

Well-behaved MCP clients back off and retry. The limit applies to the whole MCP endpoint, and /health is exempt. It is counted by each server instance separately, so treat it as the rate a client should stay under rather than an exact figure.For comparison, Query MCP allows 120 requests per minute.

Security model

If you are reviewing this server before pointing an agent at your tenant, this is the short version:

No server-held credentials. The server has no Enterspeed credential of its own. It acts only as the person who signed in, and an API key is refused as authentication.
Passthrough authorisation. Every permission decision is made by the Management API against your account, not by the MCP server.
No deletes. The server cannot issue a DELETE or a PATCH at all. Twelve writes can still change or replace existing configuration or content. See What the server can and cannot do.
Secrets are not readable. Access keys are removed from responses at the HTTP boundary, so no tool can observe one.
Every deploy and unpublish is confirmed in your client. You type the environment’s name before anything changes. A client set to auto-approve can answer for you, so keep that off on connections that can deploy.
Least privilege is yours to set. Connect with ?readOnly=true unless the agent needs to write, and with ?tenantId= to keep it on the tenants you name. See Limiting a connection.

Connecting a client

Every client connects the same way underneath: you give it the server URL, it sends you to Enterspeed in a browser window, and you sign in with your normal Enterspeed credentials. Nothing goes into your client’s configuration except the URL — with any connection parameters you want on it.

https://mcp.management.enterspeed.com/

Once you are signed in, ask your client to select the tenant you want to work on. Every session needs this, however many tenants you belong to — see Selecting a tenant.

Every client below opens the remote server directly and takes you through the sign in itself. None of them needs Node.js or a bridge program on your machine. Claude Desktop’s claude_desktop_config.json file would, because it only starts local programs, so use its Connectors screen instead. The mcp-remote bridge is only for a client that cannot open a remote server.

Agent CLIs

Claude Code
Codex
Gemini CLI
Copilot CLI

claude mcp add enterspeed-management --transport http "https://mcp.management.enterspeed.com/"

Start a session, run /mcp, and pick enterspeed-management to sign in. A browser window opens; authenticate with your Enterspeed credentials.

This installs the server for the current project only. Add --scope user to make it available in every project.

codex mcp add enterspeed-management --url "https://mcp.management.enterspeed.com/"

Codex sees that the server asks for a sign in and opens a browser window for you straight away. To sign in again later, run codex mcp login enterspeed-management.

gemini mcp add --transport http enterspeed-management "https://mcp.management.enterspeed.com/"

That writes the server into .gemini/settings.json in the current project. Add --scope user to write it into ~/.gemini/settings.json for every project instead. The equivalent by hand is this.

{
  "mcpServers": {
    "enterspeed-management": {
      "httpUrl": "https://mcp.management.enterspeed.com/"
    }
  }
}

Use httpUrl rather than a bare url. Older Gemini CLI releases, and Gemini CLI’s configuration reference, treat url as the older SSE transport, which this server does not speak.

Run /mcp inside the CLI to check the server is listed, and /mcp auth enterspeed-management to complete the sign in.

copilot mcp add --transport http enterspeed-management "https://mcp.management.enterspeed.com/"

That writes the server into ~/.copilot/mcp-config.json. The equivalent by hand is:

{
  "mcpServers": {
    "enterspeed-management": {
      "type": "http",
      "url": "https://mcp.management.enterspeed.com/",
      "tools": ["*"]
    }
  }
}

"tools": ["*"] offers every tool the server lists. The file needs the tools field, and * is what the command above writes when you give it no --tools. You can add the server interactively instead. Start copilot and run /mcp add. Copilot CLI takes you through the Enterspeed sign in in your browser the first time it connects. If /mcp shows the server as needing authentication, run /mcp auth enterspeed-management.

Web

Claude.ai
ChatGPT

On a Team or Enterprise plan only an Owner can add a custom connector. The Owner adds it under Organization settings, then Connectors. Everyone else finds it under Customize, then Connectors, and clicks Connect.

Open Customize, then Connectors.
Click + and then Add custom connector.
Give it a name, for example Enterspeed Management, and paste https://mcp.management.enterspeed.com/ as the remote MCP server URL.
Click Add, then sign in with your Enterspeed account when Claude asks.

Open Settings, then Security and login, and turn on Developer mode.
Go to ChatGPT Plugins and select the plus button.
Enter a name, and under Connection paste https://mcp.management.enterspeed.com/ as the MCP server URL.
Create the connection, then sign in with your Enterspeed account when ChatGPT asks.
In a chat, choose Developer mode from the plus menu and select the app.

OpenAI offers developer mode to Pro, Plus, Business, Enterprise and Education accounts on the web, and your account or workspace policy can decide whether you have it. If you cannot find Developer mode under Security and login, it is not available to you.

Desktop apps

Claude Desktop uses the same Connectors steps as Claude.ai, including the note for Team and Enterprise plans. The only difference is that after you click Add, the sign-in opens in a browser window.

There is no reason to use claude_desktop_config.json for this server. That file starts local programs, so it would need the mcp-remote bridge, and the result would be the same sign-in the Connectors screen gives you. On Query MCP the config-file route exists because it can carry a scoped key; this server does not accept keys.

IDEs

Cursor
VS Code

Cursor reads MCP servers from ~/.cursor/mcp.json for every project, or .cursor/mcp.json for one.

{
  "mcpServers": {
    "enterspeed-management": {
      "url": "https://mcp.management.enterspeed.com/"
    }
  }
}

A remote server is identified by url, and there is no type field to set. Cursor takes you through the Enterspeed sign in in your browser when it first connects.

VS Code’s GitHub Copilot reads MCP servers from an mcp.json file. For a single workspace, put the config at .vscode/mcp.json; for every workspace, use the MCP: Open User Configuration command from the command palette.

{
  "servers": {
    "enterspeed-management": {
      "type": "http",
      "url": "https://mcp.management.enterspeed.com/"
    }
  }
}

Reload the window (Developer: Reload Window) and open the Copilot chat pane. VS Code prompts you to sign in to Enterspeed the first time it connects, and the Enterspeed tools appear in the tool picker afterwards.

Other MCP clients

The server is client-agnostic. Any MCP client that can open a remote Streamable HTTP server and follow an OAuth sign-in works — point it at the root URL above. If a client cannot open a remote server, or does not complete the sign in, put mcp-remote in front of it. Configure it as a local command, npx -y mcp-remote "https://mcp.management.enterspeed.com/", and the bridge handles the remote connection and the sign in for the client.

Sample prompts

Use these to smoke-test a fresh connection. Orientation

Give me an overview of my Enterspeed tenant — how many sources, environments, and domains do I have, and how are they connected?

Schema investigation

List my mapping schemas and show me the current version of the one that produces my product pages. What triggers it?

Authoring loop (needs write permission in the tenant, and a connection without ?readOnly=true)

Write an index schema that indexes my article entities with title, publish date, and author. Dry-run it against a real article first, then save it as a new version.

Troubleshooting

The route for /products/red-shoe is not returning what I expect. Inspect the route and the view behind it and tell me what is wrong.

Logs

Show me any processing errors in the last 24 hours, grouped by schema.

Troubleshooting

Symptom	Likely cause	Fix
`401 Unauthorized` on every tool call	The client is not signed in, or its token has expired	Reconnect the server in your client and complete the Enterspeed sign-in. This server authenticates people, not keys — an `x-api-key` header is refused as authentication. Environment client keys belong to Query MCP.
Tenant-scoped tools fail with “no tenant is selected”	The session has not selected a tenant yet, or the selection was lost when the client reconnected or the server restarted	Ask the client to select your tenant. Every signed-in session needs this, even with only one tenant.
Writes are refused, naming read-only	The connection URL has `?readOnly=true`, or a `readOnly` value the server could not read	Expected if you set it. Reconnect without the parameter if the agent genuinely needs to write.
Writes return a permission error from Enterspeed	Your account does not have write permission in that tenant	Ask a tenant administrator for the access, or make the change in the Management App.
A tool is refused as “not offered”	The connection URL has a `?toolsets=` value that does not include that tool’s group	Add the group to `toolsets`, or use `toolsets=all`.
`deploy_not_permitted` when deploying a schema	Deploying is switched off on this server	Deploy from the Management App.
A deploy or unpublish is refused asking for a client that supports confirmations	Your client can’t show MCP confirmation prompts (elicitation)	Use a client that supports them, or deploy from the Management App.
A deploy or unpublish returns “not confirmed”	The prompt was declined, cancelled, timed out, or the environment name didn’t match	Run it again and type the environment’s name exactly.
`create_environment_client` succeeded but there is no key in the response	Working as intended — access keys are never returned through MCP	Fetch the key in the Management App.
`429` with a `Retry-After` header	More than 60 requests in a minute from your sessions	Let the client back off and retry.
Client cannot connect at all	Configured with a `/sse` path, or a proxy that buffers streamed responses	Use the root URL. Streamed responses must pass through unbuffered.

Next steps

Tools — the full tool inventory, read and write.
Query MCP — query your Enterspeed data instead of your configuration.

Schema reference

API

Integrations

Tutorials

Best practices

CLI

MCP server

Overview

Hostname

How authentication works

Selecting a tenant

What the server can and cannot do

Limiting a connection

Guardrails

Security model

Connecting a client

Agent CLIs

Web

Desktop apps

IDEs

Other MCP clients

Sample prompts

Troubleshooting

Next steps

​Hostname

​How authentication works

​Selecting a tenant

​What the server can and cannot do

​Limiting a connection

​Guardrails

​Security model

​Connecting a client

​Agent CLIs

​Web

​Desktop apps

​IDEs

​Other MCP clients

​Sample prompts

​Troubleshooting

​Next steps

Hostname

How authentication works

Selecting a tenant

What the server can and cannot do

Limiting a connection

Guardrails

Security model

Connecting a client

Agent CLIs

Web

Desktop apps

IDEs

Other MCP clients

Sample prompts

Troubleshooting

Next steps