Agents
Connect an AI agent (Claude Code, Claude Desktop, Cursor, VS Code or Gemini CLI) to the account to query configuration and events in the conversation itself. The agent accesses the account over the MCP protocol (Model Context Protocol) and operates with the permissions of the user who authorized the connection.
The feature is in testing. Functionality and usage limits may change without prior notice.
The connection is intended for internal use, by the account’s staff. Do not expose the connection to third parties nor embed it into a product offered to the account’s customers. For that use, contact support.
Connect an agent
Under Agents > Connect agent:
- Select the agent by its logo. The screen shows the instructions specific to it.
- In the agent, provide the MCP server URL and the client_id shown on the screen. If the agent asks for a client secret (
client secret), leave it blank. - When the agent requests it, authorize access to the account in the browser.
For command-line agents, the screen shows a ready-to-run terminal command in place of steps 1 and 2. Browser authorization is the same.
The screen lists the registered agents. An agent outside the list must be enabled by support before connecting.
The agent accesses one account at a time, shown on the connection screen. To query another account, connect the agent again from it.
Connected agents
Agents lists the authenticated user’s connections on the account, with the client, the authorization date and the last activity. Last activity is updated every fifteen minutes. The screen identifies the connections that were never used.
Disconnect ends the agent’s access immediately. The other connections, of the same user and of others, remain active. A disconnected agent operates again only after a new authorization.
Permissions
The agent operates on behalf of the user who authorized the connection and inherits that user’s access profile. Every call goes through two checks, and both must pass: the user’s permission and the consent granted to the agent at connection time.
| Consent | What it authorizes |
|---|---|
| Read the configuration | query event types, fields, rules, analyses, metrics, monitors and lists |
| Read the account’s data | query the events received, a metric’s measured values and the values received by a field |
| Write the configuration | change a field’s name and description |
Events contain personal data, such as emails, IP addresses, card BINs and device identifiers. Reading the account’s data is therefore a separate consent. A connection can read the account’s configuration without accessing any event.
Operations
The agent lists and details each kind of configuration, queries events and metrics and edits fields:
| Operation | Content |
|---|---|
| Account | general data of the connected account and the consents granted to the connection |
| Event types | list of types and the detail of each one, with its fields |
| Fields | values already received by a field, including by a list field with no registered items |
| Events | count of a period’s events in time blocks, with optional split by a field, and the event rows |
| Rules | list of rules and the detail of a version, with its conditions and actions |
| Analyses | list of analyses and the detail of one, with the effect of each cell and the rules that run and record it |
| Metrics | list of metrics, the detail of one (dimensions, aggregations and filters) and the values per period, with the expected band and the alerts |
| Monitors | list of monitors and the detail of one, with its criteria |
| Lists | list of the configured lists and the detail of one |
| Field editing | change of a field’s name and description |
The event query takes the same filters as Explore, including the reference to a list and the comparison between fields. The query covers a recent window, not the account’s whole history.
In clients that display MCP interfaces, a metric’s values appear as a chart in the conversation.
Each response includes the address of the corresponding article in this documentation. The agent consults the reference instead of deducing the platform’s behavior.
Typical questions:
- which rule decides a given event type;
- which metrics exist over a field;
- what a rule version does, before publishing it;
- how many events of a type arrived per status in the period.
Consents
An operation without consent does not appear to the agent. To the agent, it is the same as an operation that does not exist.
The Account operation reports the connection’s consents and the withheld operations, with the consent missing from each one. When the agent answers that something is not possible, ask it to query the account. The response shows whether the operation exists and merely was not consented to.
To release the operation, reconnect the agent granting the missing consent, as described in Connect an agent. An existing connection does not gain access without that new authorization.
Operations the user’s access profile does not allow do not appear as withheld. No reauthorization would release them, because the barrier is the user’s permission, not the agent’s.
Usage limits
Calls are limited per user and per account within a time window, with a value of its own for each operation. On exceeding the limit, the agent receives a refusal and resumes operating in the following window.