Provider REST API
InfrastructureConnect Claude to Microsoft Azure
Inventory Azure subscriptions and resources, inspect virtual machines, and read the activity log. Toolspoke puts 12 of its actions behind one MCP endpoint that Claude, Cursor, and Codex all speak.
- Connection
- Provider REST API
- Authentication
- Service principal (client credentials)
- Actions exposed
- 12
- Cost per call (typical)
- 1 credit
- Adapter
- Maintained by Toolspoke
Connected in three steps
- 1
Install Microsoft Azure
Open the marketplace in your workspace, add Microsoft Azure to the project your agents work in, and it appears on the gateway immediately.
- 2
Connect the credential
Authenticate with service principal (client credentials). Where to get one, and what it has to be able to reach, is the next section.
- 3
Point your agent at the gateway
Give your client one address, https://toolspoke.com/mcp. Claude Code takes it as a command, Claude and Claude Desktop add it as a custom connector, and Cursor, Codex and VS Code each read it from a config file of their own.
.mcp.json{
"mcpServers": {
"toolspoke": {
"type": "http",
"url": "https://toolspoke.com/mcp"
}
}
}One block covers every tool you have installed. Microsoft Azure shows up in the client as soon as your policy allows it, and so does everything else you install later.
Where the address goes, per client
- Claude Code
Run it in your project, then /mcp to sign in
claude mcp add --transport http toolspoke https://toolspoke.com/mcp- Claude and Claude Desktop
Settings, then Connectors, then Add custom connector
https://toolspoke.com/mcp- Cursor
~/.cursor/mcp.json, or .cursor/mcp.json for one project
{ "mcpServers": { "toolspoke": { "url": "https://toolspoke.com/mcp" } } }- Codex
~/.codex/config.toml
[mcp_servers.toolspoke] url = "https://toolspoke.com/mcp"- VS Code
.vscode/mcp.json, or the MCP: Add Server command
{ "servers": { "toolspoke": { "type": "http", "url": "https://toolspoke.com/mcp" } } }
What Microsoft Azure asks for
Service principal (client credentials). You provide it once, when you install the connector. Toolspoke encrypts it at rest and decrypts it only for the length of a single call, and the gateway attaches it to the outbound request itself, so it is never part of the arguments an agent sends.
- Directory (tenant) IDRequired
- Entra admin centre → App registrations → your app → Overview. It names the directory the sign-in happens in, and goes into the token endpoint address.
- 00000000-0000-0000-0000-000000000000
- Application (client) IDRequired
- Shown beside the tenant ID on the app's Overview page. This is the service principal Azure will see as the caller.
- 00000000-0000-0000-0000-000000000000
- Client secretRequired
- App registration → Certificates & secrets → New client secret. Copy the Value, not the Secret ID - the Value is shown once and never again. Give the service principal its RBAC role separately, under the subscription's Access control (IAM); Reader covers every read this connector makes.
- Paste the secret value
- Default subscription IDOptional
- Used whenever an action is called without an explicit subscription_id. Leave it blank to require one on every call; list_subscriptions shows what this service principal can see.
- 00000000-0000-0000-0000-000000000000
What Claude can do in Microsoft Azure
12 actions, each one declared and named by the connector rather than discovered at runtime. A workspace policy grants a person all of them, a hand-picked selection, everything on the read side, everything on the write side, or none.
- Reads
- 9Reads
- Writes
- 1Writes
- Destructive
- 2Destructive
Reads
9Fetches data and changes nothing.
list_subscriptionsList every subscription this service principal can see, with each one's id, display name and state. Call it first when you do not know which subscription to act on, and to find out whether the connection reaches anything at all - an empty list means the app registration exists but has been given no RBAC role on any subscription. The subscriptionId returned here is what every other operation's subscription_id argument takes.
list_resource_groupsList the resource groups in one subscription, with each group's name, location and tags. Resource groups are how Azure work is usually organised, so this is the natural second call after list_subscriptions and the way to find the resource_group name that the virtual machine operations need. Use list_resources instead when the question is about a kind of resource rather than about where things are grouped.
list_resourcesList resources across a whole subscription, of any type, with each one's full ARM id, name, type, location and tags. This is the general inventory call and the one to reach for when the question is "what is running here" or "where is the thing called X" - narrow it with `filter`, which is the only practical way to keep the answer small in a busy subscription. The `id` it returns is the fully qualified resource id that get_resource takes. For virtual machines specifically, list_virtual_machines returns their configuration rather than this generic envelope.
get_resourceFetch one resource of any type by its fully qualified ARM id, returning the provider's own properties for it rather than the generic summary a listing gives. Ids come from list_resources and look like /subscriptions/{guid}/resourceGroups/prod-rg/providers/Microsoft.Web/sites/my-app. IMPORTANT: api_version here belongs to the resource's own provider, not to ARM - Microsoft.Web, Microsoft.Compute and Microsoft.Storage each version independently, and a version the provider does not recognise is rejected outright. Read the `type` from list_resources, then pass the api-version documented for that type.
list_virtual_machinesList every virtual machine in a subscription with its size, image, OS disk, network interfaces and provisioning state. This returns the machine's configuration, which is what it was built as - it does not say whether the machine is running right now. For that, call get_vm_instance_view with the resource group and name from here. The `id` of each machine is also a valid get_resource id.
get_vm_instance_viewReport one virtual machine's run-time state: power state (running, stopped, deallocated), provisioning state, agent and extension health, boot diagnostics and disk status. This is the operation that answers "is it up" - list_virtual_machines describes how a machine is configured and says nothing about whether it is currently running. Call it before and after start_vm, restart_vm or deallocate_vm, since all three return as soon as Azure accepts the request rather than when the machine has finished changing state. Resource group and machine name come from list_virtual_machines.
list_storage_accountsList the storage accounts in a subscription with their kind, SKU, replication, endpoints and network access settings. Useful for finding where blob data lives and for auditing exposure - allowBlobPublicAccess, minimumTlsVersion and publicNetworkAccess all come back here. Access keys are deliberately not returned by this endpoint and this connector exposes no operation that returns them.
query_activity_logsRead the subscription's activity log - the control-plane audit trail of who changed what and whether it worked. This is the operation to reach for after an unexplained change: it names the caller, the operation, the resource and the outcome. `filter` is REQUIRED and Azure accepts only a handful of exact shapes, always starting with a time range: "eventTimestamp ge '2026-08-20T00:00:00Z' and eventTimestamp le '2026-08-21T00:00:00Z'", optionally followed by exactly one of `and resourceGroupName eq '…'`, `and resourceUri eq '…'`, `and resourceProvider eq '…'` or `and correlationId eq '…'`. No other syntax is accepted, and a malformed filter is rejected rather than ignored. The log holds 90 days.
list_app_servicesList the App Service web apps and function apps in a subscription, with each one's state, default hostname, hosting plan, runtime stack and HTTPS settings. Use it to find which app serves a hostname, or to check whether an app is stopped. Deployment, restarts and configuration changes are not exposed here - this connector reads App Service and does not operate it.
Writes
1Creates or updates something on the other side.
start_vmStart a stopped or deallocated virtual machine. Classified a write rather than destructive because it brings capacity back rather than taking it away, and calling it on a machine that is already running is harmless. It does cost money the moment the machine boots. Azure answers 202 Accepted immediately and starts the machine in the background, so confirm with get_vm_instance_view rather than treating the response as proof it is up.
Destructive
2Deletes or permanently alters something. Worth granting on purpose.
restart_vmRestart a running virtual machine. Everything on it goes down and comes back, which is visible to whatever depends on that machine and is not undone by calling start_vm afterwards - so this is classified destructive, out of reach of a read-only or write grant, and an install has to grant full access or name it explicitly. Azure answers 202 Accepted at once; watch the power state with get_vm_instance_view to see the restart finish.
deallocate_vmShut a virtual machine down and release its compute, which is the operation that actually stops the bill for it - an in-guest shutdown leaves the machine allocated and still charged. The machine goes down, its dynamic IP is released unless it is a static one, and local temporary disks are lost, so this is classified destructive and cannot be reached by a read-only or write grant. start_vm brings it back, but not to the same ephemeral state. Azure answers 202 Accepted at once; confirm the deallocated power state with get_vm_instance_view.
What it will not do
Enforced by the gateway rather than left to convention, which is why each of these can be stated flatly.
- It cannot call anything else
- The 12 actions above are the whole of it. A call to any other name is refused before it reaches Microsoft Azure rather than forwarded on, and connecting your account does not add to the list: it is fixed by the connector, not discovered at run time.
- It reaches no further than your credential
- Toolspoke holds no access to Microsoft Azure of its own. Every call carries the credential you stored and nothing besides, so whatever that credential cannot reach, this connector cannot reach either.
- It never hears from Microsoft Azure
- Nothing is pushed to it. There is no webhook, no subscription and no polling, so this connector cannot notice by itself that something changed in Microsoft Azure. An agent has to ask.
- It does not smooth over provider limits
- Toolspoke does not retry, queue or back off around Microsoft Azure's own rate limits. A call that Microsoft Azure refuses comes back to the agent as a failed call.
Before you connect it
What can Claude do in Microsoft Azure?
12 named actions: 9 that only read, 1 that write and 2 that delete or permanently alter something. They include list_subscriptions, list_resource_groups and list_resources. Nothing outside that list is reachable: the connector declares each operation by name rather than proxying whatever an agent asks for.
What credentials does the Microsoft Azure connector need?
Service principal (client credentials). The connector asks for directory (tenant) id, application (client) id and client secret, and optionally default subscription id. Values are encrypted at rest and attached to the outbound request by the gateway, so they are never part of the arguments an agent sends and never reach the audit log.
Does the Microsoft Azure connector work with Cursor and Codex, or only Claude?
Any client that speaks MCP, and every one of them gets the same 12 actions. There is a single address, https://toolspoke.com/mcp. Claude Code adds it with claude mcp add --transport http, Claude and Claude Desktop take it as a custom connector in settings, Cursor reads it from .cursor/mcp.json, Codex from ~/.codex/config.toml, and VS Code from .vscode/mcp.json. Each of them signs in to the gateway itself, so there is no key to paste.
What does the Microsoft Azure connector not do?
The 12 actions above are the whole of it. A call to any other name is refused before it reaches Microsoft Azure rather than forwarded on, and connecting your account does not add to the list: it is fixed by the connector, not discovered at run time. Toolspoke holds no access to Microsoft Azure of its own. Every call carries the credential you stored and nothing besides, so whatever that credential cannot reach, this connector cannot reach either. Nothing is pushed to it. There is no webhook, no subscription and no polling, so this connector cannot notice by itself that something changed in Microsoft Azure. An agent has to ask. Toolspoke does not retry, queue or back off around Microsoft Azure's own rate limits. A call that Microsoft Azure refuses comes back to the agent as a failed call.
Can I limit which actions an agent can call?
Yes, in two places. The project switches Microsoft Azure's actions on and off one at a time, for everyone in the project at once, and the screen groups them by read, write and destructive so turning off everything that deletes is one click. An individual agent key can then be narrowed further, to particular toolkits in a project and to particular actions in a toolkit. Whatever it was granted, a key never reaches a project its owner cannot.
What gets recorded when an agent calls Microsoft Azure?
Every attempt, with the agent that made it and the person that agent belongs to, the full request payload, the response payload, the status, the duration, and the credits spent. Values whose key names a secret are masked out before the record is shown to anyone. An operation the connector marks as not retained never has its response body written at all, so the gateway keeps no second copy of what was read.