Sensos MCP server

Instead of opening Sync, filtering a list, and exporting it, ask: "Which shipments to Frankfurt went over temperature last month?"
Your assistant runs the lookups it needs, combines them, and answers in one pass.

The Sensos MCP server uses the Model Context Protocol, an open standard that lets AI assistants query external systems.

See Supported clients for which AI clients can connect today.

Two facts govern everything else on this page:

• The server is read-only. It cannot create, change, or delete anything in Sensos, and there is no setting that turns writing on.
• It sees only what you see. Your existing Sensos permissions decide every result, and they are checked on every request.

Connection details

You do not need an API key, and there is nothing for you to register.
Your AI client identifies itself with a client metadata document (CIMD): a public page the client's maker hosts that describes the client, which Sensos reviews and allow-lists once for everyone.
Connecting then simply sends you to the Sensos sign-in page to authorize the connection.

Sensos deliberately does not support anonymous dynamic client registration.
A client Sensos has not allow-listed is refused before sign-in even starts, so only known clients can ask your permission to connect.

Supported clients

• Claude (web and desktop) is verified by Sensos and listed in the Claude connectors directory. See Connect Sensos to Claude for the steps.
• Claude Code is verified by Sensos: claude mcp add --transport http sensos https://mcp.sensos.io/mcp
• Other MCP clients must support CIMD and be allow-listed by Sensos before they can connect; clients that only support dynamic client registration cannot connect. Client makers can reach out to [email protected] to have their CIMD added.

What you can do

Shipments. Search by status, carrier, route, alert type, Set, location, custom property, or free text.
Then read one shipment in full: where it is, which alerts fired and why, milestones, carrier and airway-bill events, the plan, the items on board, and the latest notes.

Alerts. Search alerts across shipments and assets, open and closed, newest first.
Each alert carries its condition, category, severity, lifecycle status, fired and closed times, and the measured values that tripped it.

Assets. Find reusable assets by type, home location, tag, alert state, or custom property, and read one in full.

Labels. Look up a physical Sensos Label by identifier or free text: status, battery, reporting history, and which shipments or assets it has been paired to and when.

Locations, Sets, and Items. Look up your saved locations, the Sets that define monitoring and alert thresholds, and the part numbers and serial numbers your shipments carry.

Analytics. Ask for aggregated counts and trends instead of paging through records: completions by week, alerts by condition, durations, label usage, estimated emissions.

Two boundaries, so an answer never looks like a bug:

• The server answers questions you ask.
  It cannot watch your data in the background or notify you when something changes.
  Alerts still reach you through Sensos Sync and your normal notification settings.
• Shipment and asset detail return the journey as a sequence of events plus a report count, not the raw position-by-position device trace.
  That stays in Sensos Sync.

Permissions and data access

Read-only, enforced rather than assumed

Every tool the server publishes is marked read-only, and Sensos publishes no write tools at all.
The connection holds a single read permission, every tool requires it, and Sensos re-checks that permission when a tool runs rather than trusting that a write tool was simply never shown.

Your Sensos permissions decide everything

Connecting an AI assistant does not widen your access.
The server cannot show you anything you could not open in Sensos Sync yourself.

Sensos decides this on every request, not once at connection time.
If your access changes in Sensos, the next question you ask reflects the change.

Every tenant and account you belong to, in one connection

The connection is not pinned to one tenant or one account.
Your assistant begins by asking Sensos which tenants you belong to and which accounts each contains, then names one tenant and one account on each lookup.

So you can ask about several accounts in a single conversation without disconnecting and reconnecting.
Each lookup covers one tenant and one account, so a question spanning several accounts becomes several lookups, which your assistant handles for you.

Naming a tenant is not the same as being allowed into it.
Sensos re-confirms your membership of a tenant whenever it issues the internal credential for that tenant and refuses if you are not a member, so a wrong or invented tenant name cannot become access.
Every data request is then checked again, live, against your current permissions.

Authentication

You sign in to Sensos directly, and your AI client never sees your Sensos password.

Your client receives a token that works only against the Sensos MCP server.
Internally, Sensos exchanges that token for a separate short-lived credential to read your data.
The second credential is never given to your client.

Available tools

Your assistant picks and combines these on its own, based on your question.
There are 14, and all of them only read.

Vocabulary. Saved locations appear as Locations in Sync and are sometimes called points of interest.
Labels are the physical Sensos tracking devices.
Sets are the configuration templates that define monitoring rules and alert thresholds.

Two gaps in alert coverage. Reading one shipment or asset includes its closed and Informational alerts but excludes Low Battery and Connectivity alerts; ask for those through alert search instead.
Alert search never returns Informational alerts.

Example prompts

Naming an account, a date range, or an identifier makes the answer much better.

• "Which Sensos tenants and accounts can I reach?"
• "Show in-transit shipments in [ACCOUNT NAME] with open temperature alerts, and tell me which ones need attention first."
• "Give me the full picture for shipment [SHIPMENT NAME OR ID]: where it is, every alert that fired and why, its milestones, and the latest notes."
• "Find assets in [ACCOUNT NAME] currently in an alert state, grouped by asset type."
• "Check Label [LABEL ID]: battery, last report, and what it is paired to right now."
• "What temperature thresholds apply to shipment [SHIPMENT NAME OR ID]? Look up its Set and tell me the actual limits."
• "Show shipment completions by week for the last 90 days in [ACCOUNT NAME], and flag any week that looks unusual."

And one longer investigation, which is where the server earns its keep:

"Shipment [SHIPMENT NAME OR ID] arrived late and the customer is claiming a temperature excursion. Pull the shipment, its alerts with measured values and timestamps, its milestones and carrier events, and the thresholds from its Set. Then tell me whether the excursion is real against those thresholds, when it happened relative to the delay, and what the notes say. Say clearly which parts are supported by the data and which are inference."

That last sentence is worth building into any prompt you reuse.
Asking your assistant to separate what the data shows from what it is inferring keeps a plausible answer from being mistaken for a sourced one.

Limits and fair use

Requests to the server are rate limited to protect the service.
An assistant that sends a large burst of lookups may have some of them briefly refused; asking again after a moment succeeds.
Sensos does not publish the exact limits, and they may change.

Search results are paged, and large detail collections are capped rather than returned whole.
When a result is truncated, the response says so, and your assistant can ask for more pages.

For Sensos administrators

• There is no separate credential to manage.
  Access rides entirely on the user's Sensos account.
  Removing a user's Sensos access, or their membership of a tenant, stops their MCP access to that data from their very next request, because Sensos checks the user's current permissions live on every data request rather than trusting anything issued earlier.
  The same freshness applies in the other direction: access granted to a user, including time-limited support access, is usable right away.
  The one change that can lag is deactivating an entire tenant, which propagates within a few minutes.
• Requests are attributed.
  Sensos records MCP requests in its operational telemetry, attributed to the connector, for security and operations purposes.
  Contact Sensos support if you need an access review.
• To manage connector use for your organization, for example to prevent connector use while keeping Sync access, contact Sensos support.

Security and privacy

What leaves Sensos. Only the data needed to answer the question you asked, retrieved with your own permissions, and returned to your assistant's conversation.

What stays in Sensos. Your credentials, the internal credential used to read your data, and everything you did not ask about.

What your AI provider handles. Your questions and the returned data become part of your conversation with your AI assistant, handled by that provider under the terms that apply to your account or organization.
For Claude specifically, see Connect Sensos to Claude.

What the server never touches. It requests no access to your assistant's memory, chat history, conversation summaries, or files.

Using Sensos alongside other connectors. In a conversation with several connectors, content returned by one tool can influence what your assistant asks the others, and Sensos data can flow onward to other connected tools in the same conversation.
The Sensos server being read-only means it cannot be made to change anything, but treat the conversation itself as the boundary: connect Sensos only in conversations and organizations approved to receive your Sensos data.

What Sensos retains. Sensos keeps records of connector sign-ins and authorizations, and operational telemetry about each request: which tool was called, with which parameters, when, and how it performed.
That telemetry includes the search parameters your assistant used, but not the data returned to it.

Sensos handles connector authentication, the Sensos data you request, and operational telemetry, including how long each is retained, as described in the Sensos privacy policy.

Disconnect or revoke access

To stop using the server, disconnect or remove it in your AI client; for Claude, the steps are on Connect Sensos to Claude.
No further requests can be made from that connection.

Data already returned into a conversation stays in that conversation, governed by the retention and deletion controls of your AI provider account or organization.
Delete the conversation if you need the content gone.

Because access is re-checked against Sensos on every request, removing a user's Sensos access stops their connector from their very next request.
There is nothing separate to revoke on the Sensos side.

Get help

For help with the server or with Sensos data, contact [email protected].
Include roughly when the problem happened, the question you asked, which AI client you used, and any reference ID shown in an error.
Never send passwords, tokens, or confidential shipment data by email.

To report a suspected security vulnerability, email [email protected] and say clearly that it is a security report.
Do not include live credentials or unnecessary customer data in a first report.