Technical reference
Reference for the nine tools provided by SilentLog Analytics MCP and their principal parameters, written for engineers evaluating the service. For an overview, see the service page (日本語).
Last updated: 30 August 2026
How you connect
Provided as an MCP (Model Context Protocol) Streamable HTTP endpoint. Authentication is by JWT (RS256) or API key. Any MCP-compliant client can connect, including AI assistants and applications you build yourself. Testing was carried out against the Claude API.
Tool structure
The nine tools sit in four layers. The AI assistant selects and combines whichever ones your question requires.
| Layer | Tools | Role |
|---|---|---|
| Geography | maps_address_search maps_geocoding maps_schema_list | Resolve coordinates from names and addresses; find what is near a point |
| Retrieval | silent_log_search profile_search | Retrieve behavioural data; search estimated attributes |
| Combined | silent_log_analytics | Retrieve and aggregate in one call |
| Computation | analytics_statistics analytics_clustering analytics_histogram | Process data you have already retrieved |
Retrieving behavioural data silent_log_search
Retrieve behavioural data for a given location and period.
| Parameter | Type | Default | Notes |
|---|---|---|---|
| start_datetime end_datetime | string | — | Period to analyse, ISO 8601. Maximum 31 days. |
| center_lat / center_lon | number | — | Centre point of the search |
| address / facility_name | string | — | Specify the point by address or facility name instead of coordinates |
| radius_km | number | 1.0 | Search radius. Maximum 10.0 |
| activity_filters | array | all | sty staying / wlk walking / trp travelling |
| move_type_filters | array | — | walking bicycle car train bus and others |
| null_policy | string | exclude | How undetermined records are treated. Use include to count them |
| limit | integer | 100 | Number of records. Maximum 10000 |
| timezone | string | Asia/Tokyo | IANA time zone name |
| exact_total | boolean | false | true returns an exact total count |
| cursor | string | — | Fetch the next page |
Limits on period and area
A single request covers at most 31 days; anything longer returns an error, so long periods are analysed in parts. The radius is capped at 10 km and larger values are silently clamped. A location is always required, so aggregating the whole country in one call is not possible.
Retrieval and analysis combined silent_log_analytics
Performs retrieval and aggregation together. In addition to the parameters above, the following apply.
| Parameter | Type | Default | Notes |
|---|---|---|---|
| group_by | string | — | day / hour (0–23) / dow (day of week). Switches the response to bucketed aggregates |
| analysis_types | array | statistics, clustering | statistics clustering histogram |
| sampling | string | by distance | random or time_stratified to change how records are drawn |
| clustering_method | string | dbscan | kmeans dbscan hierarchical |
| compact | boolean | false | true returns aggregates only |
Use group_by for time-based aggregates
Setting
group_by changes the response into per-bucket aggregates. This path is not subject to the record limit, so it is the reliable way to obtain counts by day, hour or day of week. Note that buckets with no matching records are omitted entirely, so do not assume a contiguous sequence.
Estimated attributes profile_search
Search estimated data on visitors, including place of residence, place of work and other attributes.
| Parameter | Type | Default | Notes |
|---|---|---|---|
| query | string | — | Semantic search query. Either this or user_id is required |
| month | string | 202410 | Target month in YYYYMM format |
| top_k | integer | 10 | Number of results |
| gender | string | — | 男性 (male) / 女性 (female) |
| age_min / age_max | integer | — | Age range |
| residence_prefecture workplace_prefecture | array | — | Prefecture names. Multiple values are treated as OR |
Mind the reference date
Attribute estimates are a snapshot from a fixed point in time and are not refreshed monthly. Because this differs from the timing of the behavioural data, please consult us before presenting the two together.
Maps and address lookup maps_*
| Tool | Principal parameters | Defaults | Description |
|---|---|---|---|
| maps_address_search | query / facility_name / address / region | country=JP, limit=10 | Resolve coordinates from a facility name or address |
| maps_geocoding | latitude+longitude or coordinates | radius_km=1.0, limit=50 | Find nearby facilities from coordinates. Up to 100 coordinates per call |
| maps_schema_list | filter_theme | all themes | List the available map data themes |
Statistical processing analytics_*
These are computation-only tools. They operate on data you pass in and do not retrieve anything themselves.
| Tool | Parameters | Description |
|---|---|---|
| analytics_statistics | datarequired, columns, statistics | Mean, median, standard deviation, quantiles and more |
| analytics_clustering | datarequired, columns, method, n_clusters | Cluster locations or behavioural patterns. Up to 256 columns / 500,000 cells |
| analytics_histogram | datarequired, columnrequired, plot_type, bins | Produce distributions |
What the response tells you
The response reports whether your conditions were applied as specified.
| Field | Meaning |
|---|---|
| total_distinct_users | Unique user count. Use this figure when reporting population size |
| total_display | Total records. Set exact_total to true for an exact figure |
| clamp_notices | Recorded when a value you supplied was clamped to a limit |
| filter_metadata | The filters actually applied |
| truncated | Whether results were cut short by the record limit |
Further documentation
On adoption we provide an API usage guide, deployment instructions and an operations guide. We will also advise on the connection method that suits your environment, so please get in touch.