MongoDB Atlas
Manage MongoDB Atlas: projects and clusters, database users and IP access lists, alerts and events — the control plane, not the data.
About
MongoDB Atlas ships in the w6w first-party pack. It declares 19 actions, 3 health checks, and the host runs its code in a sandbox that never sees the credential.
- App id
io.w6w.mongodb-atlas- Version
- 0.1.2
- Author
- w6w
- Licence
- MIT
- Categories
- Databases · DevOps & Infrastructure
Overview
MongoDB Atlas is the managed hosting layer around a MongoDB deployment, and this app manages its control plane — projects, clusters, database users and network access — not the data inside a database, which needs a driver speaking the wire protocol directly. Create, scale, pause and delete clusters, mint or revoke database user credentials, and open or close the IP access list entries that gate who can connect at all.
Visibility comes from reading alerts, change events and the running processes behind a cluster, including which node is currently primary. Cluster actions default toward safety where Atlas doesn’t: creation requires an explicit instance size and turns on termination protection, deletion keeps backup snapshots unless told otherwise, and opening network access to everyone requires an explanatory comment.
Useful for provisioning a cluster as part of an environment setup, rotating a database credential on a schedule, pausing non-production clusters overnight to cut cost, or reacting to Atlas’s own alert and event stream instead of watching the console.
Build with MongoDB Atlas
Three routes to the same 19 actions. The Workflow tab is generated from MongoDB Atlas's own manifest and carries its real ids, so it is copy-pasteable; the Code and CLI examples are the same call for any action on any app, so every app-specific value in them is a blank you fill in.
Add an IP access entry
access-list-add Let an address or CIDR block reach the project's clusters — ALL of them, present and future. `0.0.0.0/0` is gated, and an entry with no expiry is one nothing will remove.
Delete an IP access entry
access-list-delete Stop an address reaching the project's clusters, effective for new connections at once. The CIDR slash must be encoded into the path, which is why a hand-rolled call 404s.
List IP access entries
access-list-get The project's IP access list — the perimeter for every cluster in it. Flags `0.0.0.0/0`, which is the whole internet, and counts entries with an expiry date.
List alerts
alert-list A project's alerts, OPEN by default because unfiltered the list is mostly resolved history. `TRACKING` means the condition is met and inside its notification delay.
Create a cluster
cluster-create Provision a dedicated cluster, which BILLS PER HOUR from creation. Termination protection is on by default here, against Atlas's own default. Creation takes minutes and the cluster refuses changes until it is IDLE.
Delete a cluster
cluster-delete Destroy a cluster and its data. Atlas REFUSES this while termination protection is on. Backups go too unless retained, leaving no copy of the data anywhere.
Get a cluster
cluster-get One cluster, with its connection string, state and whether termination protection is on. The connection string carries NO credentials — a database user and an access-list entry are separate.
List clusters
cluster-list A project's dedicated clusters. FLEX clusters are a separate endpoint and are NOT in here — an audit built on this list silently misses them.
Pause or resume a cluster
cluster-pause Stop or restart a cluster's compute billing, keeping its data. Atlas RESUMES a paused cluster after 30 days by itself, and refuses to re-pause one for 60 minutes after it resumes.
Update a cluster
cluster-update Change a cluster's size, backup or protection settings. Scaling rebuilds nodes over tens of minutes with a primary election in the middle, and the cluster refuses further changes until it is IDLE again.
Create or update a database user
database-user-create Create a database credential, or rotate an existing one's password. Passwords are WRITE-ONLY — nothing returns them — and a user with no scopes can reach every cluster in the project.
Delete a database user
database-user-delete Revoke a database credential. EXISTING CONNECTIONS ARE NOT CLOSED — this stops the next authentication, so a running application keeps working until something makes it reconnect.
List database users
database-user-list The credentials applications connect with — not Atlas accounts. A user with no `scopes` can reach EVERY cluster in the project, present and future, which is the default.
List project events
event-list The project's audit trail — who created, resized or deleted what, and when. Events caused by automation are attributed to the SERVICE ACCOUNT, so one account per automation is what makes this readable.
List flex clusters
flex-cluster-list The flex tier, which `cluster-list` does NOT return — a separate endpoint that exists only from resource version 2024-11-13. Between the two you have the project's real inventory.
List organizations
organization-list The organisations this service account can reach. An EMPTY result means the account exists and has been granted no role — it authenticates perfectly and sees nothing.
List cluster processes
process-list The individual nodes behind a project's clusters, with which is PRIMARY — the only place this API says so. Their ids are what the per-node metrics endpoints take.
Get a project
project-get One project's details, including how many clusters it holds. A malformed id would answer 401 rather than 400, so the shape is checked here first.
List projects
project-list The projects this account can reach — `groups` in every path, `projects` in the console. This is how a project NAME becomes the id everything else needs.
A workflow step names the app and the action, and the editor fills in the
connection when you pick one. This is the Step shape from the
workflow spec, carrying MongoDB Atlas's real ids.
{
"manifestVersion": "2",
"name": "mongodb-atlas-example",
"steps": [
{
"id": "access-list-add",
"uses": {
"app": "io.w6w.mongodb-atlas",
"action": "access-list-add",
"connection": "conn_YOUR_CONNECTION_ID"
},
"with": {
"projectId": "<projectId>",
"value": "<value>",
"comment": "<comment>"
}
}
]
}Here are some of the things you can do
-
Add an IP access entry
performaccess-list-add -
List IP access entries
readaccess-list-get -
List alerts
readalert-list -
Create a cluster
performcluster-create -
Get a cluster
readcluster-get
+14 more actions available
Every app-specific value here is a blank you have to fill in. An
app action is reached through the connection that authenticates it, so the
address is a connection id, not the app id — and connections belong to your account,
so a public page cannot know yours. Create one for MongoDB Atlas, then fill in
the three blanks: conn_YOUR_CONNECTION_ID, the action key, and the
parameters that action declares. The call itself is real — the shape is transcribed
from the studio's own snippet builder, which prints the same kind of blanks — but
nothing in it is specific to MongoDB Atlas. The Workflow tab is where this app's
real ids are.
npm install @w6w/sdkyarn add @w6w/sdkpnpm add @w6w/sdkdeno add npm:@w6w/sdkimport { W6wClient, isActionRun } from "@w6w/sdk";
// Reads W6W_BASE_URL and W6W_TOKEN from the environment when omitted.
const client = new W6wClient();
const envelope = await client.run({
urn: "conn_YOUR_CONNECTION_ID",
action: "access-list-add",
payload: {
projectId: "<value>",
value: "<value>",
comment: "<value>",
// deleteAfterDate: "<value>",
// confirmOpenToInternet: "<value>",
},
});
if (isActionRun(envelope)) console.log(envelope.value);npm install -g @w6w/cliw6w run conn_YOUR_CONNECTION_ID --action access-list-add --payload '{"projectId":"<value>","value":"<value>","comment":"<value>"}'Give an AI agent MongoDB Atlas — without giving it MongoDB Atlas's credentials. One MCP endpoint exposes every app, function and workflow the caller is entitled to, as tools it can discover and run. Access is granted per team while we onboard.
One tool call{
"name": "w6w_invoke",
"arguments": {
"ref": "app:io.w6w.mongodb-atlas#access-list-add",
"input": {
"projectId": "<projectId>",
"value": "<value>",
"comment": "<comment>"
}
}
}
Every tool names its target with a single ref. The
app: form above doesn't name a connection at all — the
host resolves which of the caller's MongoDB Atlas connections to sign
with, and refuses rather than guesses when the answer is ambiguous.
What the agent gets
Credentials it can't read
The token is attached host-side, at the moment of the call. It is never a tool argument, never in the model's context, and never in a transcript — so a prompt injection has nothing to exfiltrate.
A tool surface scoped to the caller
Tools are derived per end user from what that person has actually connected and is entitled to — not one shared bot identity carrying the union of everyone's access.
A durable workflow in one call
Multi-step work runs on the workflow engine and returns a run handle the agent can poll — retries, branching and state survive the conversation that started them.
Health-aware discovery
MongoDB Atlas's declared health checks are on the surface too, so an agent can tell "the vendor is down" from "your credential expired" before it burns a retry on either.
The MCP surface is part of the hosted platform. MongoDB Atlas itself is MIT, and the runtime that executes it is source-available (FSL).
Health checks
MongoDB Atlas declares its own checks, so its health is a property of the app rather than something the host guesses at.
MongoDB Cloud status
Reads status.mongodb.com for the `MongoDB Cloud` component — the console and this Administration API. It says NOTHING about whether your clusters are reachable: a driver talks to them directly over the wire protocol, which does not pass through this API at all.
Token accepted
Reads the organisation list with this connection's token. Distinguishes a rejected token from one that works and can see NOTHING — a service account with no role granted authenticates perfectly and is useless, and nothing else reports that.
Rate-limit headroom
Not checkable. Atlas documents 100 requests per minute PER PROJECT and publishes no rate-limit header at all, so there is nothing to sample — and the constraint that actually bites is a cluster refusing changes while it is not IDLE.