:hedgehog: PostHog is the leading platform for building self-driving products. Our developer tools – AI observability, analytics, session replay, flags, experiments, error tracking, logs, and more – capture all the context agents need to diagnose problems, uncover opportunities, and ship fixes. Steer it all from Slack, web, desktop, or the MCP.
Python ab-testingai-analyticsanalyticscdpdata-warehouseexperimentsfeature-flagsjavascriptproduct-analyticspythonreactsession-replaysurveystypescriptweb-analytics
import { DataTableNode, EventsQuery, NodeKind } from '~/queries/schema/schema-general'
import { AnyPropertyFilter, PropertyFilterType, PropertyOperator } from '~/types'
import {
DEFAULT_MCP_ACTIVITY_QUERY,
MCP_ACTIVITY_COLUMNS,
MCP_ACTIVITY_MAX_ROWS,
MCP_ACTIVITY_PAGE_SIZE,
MCP_RECENT_TOOL_CALLS_LIMIT,
buildRecentToolCallsQuery,
withFailedCallsOnly,
} from './toolCallFeedQuery'
describe('tool call feed queries', () => {
it('starts the expandable MCP activity feed at 100 rows with room to load more', () => {
expect(DEFAULT_MCP_ACTIVITY_QUERY).toMatchObject({
embedded: false,
expandable: true,
showCount: true,
showDateRange: true,
showPropertyFilter: expect.any(Array),
source: {
events: ['$mcp_tool_call'],
limit: 100,
orderBy: ['timestamp DESC'],
},
})
expect(MCP_ACTIVITY_PAGE_SIZE).toBe(100)
expect(MCP_ACTIVITY_MAX_ROWS).toBeGreaterThan(MCP_ACTIVITY_PAGE_SIZE)
expect(MCP_ACTIVITY_COLUMNS).toContain('*')
expect(MCP_ACTIVITY_COLUMNS.find((column) => column.endsWith('-- Tool'))).toContain('$mcp_exec_tool_call_name')
expect(MCP_ACTIVITY_COLUMNS.find((column) => column.endsWith('-- Model'))).toContain('$mcp_llm_model')
})
it('narrows the feed to failed calls without dropping other filters or stacking itself', () => {
const toolFilter: AnyPropertyFilter = {
key: '$mcp_tool_name',
value: ['search'],
operator: PropertyOperator.Exact,
type: PropertyFilterType.Event,
}
// Same key on a person property is a different filter and must survive.
const personFilter: AnyPropertyFilter = {
key: '$mcp_is_error',
value: ['true'],
operator: PropertyOperator.Exact,
type: PropertyFilterType.Person,
}
const query: DataTableNode = {
...DEFAULT_MCP_ACTIVITY_QUERY,
source: { ...(DEFAULT_MCP_ACTIVITY_QUERY.source as EventsQuery), properties: [toolFilter, personFilter] },
}
const once = withFailedCallsOnly(query)
const twice = withFailedCallsOnly(once)
expect(once.source.kind).toBe(NodeKind.EventsQuery)
expect(twice.source).toMatchObject({
properties: [toolFilter, personFilter, { key: '$mcp_is_error', value: ['true', '1'] }],
})
})
it('scopes the dashboard glance to the dashboard filters and drops the toolbar', () => {
const toolFilter: AnyPropertyFilter = {
key: '$mcp_tool_name',
value: ['search'],
operator: PropertyOperator.Exact,
type: PropertyFilterType.Event,
}
const query = buildRecentToolCallsQuery({
dateRange: { date_from: '-7d', date_to: '-1d' },
properties: [toolFilter],
filterTestAccounts: true,
})
expect(query).toMatchObject({
embedded: false,
showDateRange: false,
showPropertyFilter: false,
showReload: false,
source: {
select: MCP_ACTIVITY_COLUMNS,
after: '-7d',
before: '-1d',
properties: [toolFilter],
filterTestAccounts: true,
limit: MCP_RECENT_TOOL_CALLS_LIMIT,
},
})
})
})