The Confidence MCP servers are experimental. Tools and their parameters can change.
Endpoint and Authentication
Endpoint:https://mcp.confidence.dev/mcp/flags. Transport: streamable HTTP. Authentication: sign in to Confidence from your AI assistant. The server registers 56 tools. Where a tool returns one page of results, pass the returned nextPageToken as pageToken to get the next page.
Clients
| Tool | Description | Key parameters |
|---|---|---|
listClients | Lists Confidence clients with their display name and resource name (clients/<id>). | pageToken |
createClient | Creates a client for resolving flags. The client type is Frontend or Backend. Returns the display name, resource name, and client secret. | displayName, clientType |
getClientSecret | Returns the client secret that applications use to authenticate when they resolve flags. | clientName, credentialId |
Flags
| Tool | Description | Key parameters |
|---|---|---|
listFlags | Lists flags. In list mode, narrow results with state, owner, and filter. In search mode, pass query for full-text search ordered by relevance. Search mode ignores the other filters. | query, state, owner, filter, pageToken |
getFlag | Returns a flag. Set summary=false for detailed rule information, including rollout percentages and timestamps. | flagName, summary |
createFlag | Creates a flag with a name, description, schema, and variants. | flagName, description, clientName, schemaObject, variants |
batchCreateFlags | Creates many flags in one call, using the same definition format as createFlag. Skips flags that already exist and returns created, existing, and failed counts. | clientName, flags, labels |
addFlagVariant | Adds a variant to an existing flag. | flagName, variantName, value |
updateFlagSchema | Updates a flag’s schema. You can add properties when variants exist. The backend rejects incompatible changes, such as removing a property that variants use. | flagName, schemaObject |
addFlagToClient | Enables a flag for a client. | flagName, clientName |
removeFlagFromClient | Disables a flag for a client. | flagName, clientName |
archiveFlag | Archives a flag. An archived flag can’t resolve and isn’t included in batch resolves. | flagName |
unarchiveFlag | Restores an archived flag so it resolves again. | flagName |
updateFlagOwner | Changes the owner of a flag. The owner must be an exact identity name from getIdentityInfo or searchIdentities. | flagName, owner |
batchUpdateFlagOwner | Changes the owner of many flags. Select flags with a comma-separated flagNames list, or with currentOwner to move every flag that identity owns. | owner, flagNames, currentOwner, state |
getFlagHistory | Returns the audit log of changes to a flag within a required date range: what changed, when, and by whom. | flagName, fromDate, toDate, pageToken |
analyzeFlagUsage | Finds flags on a client that nothing uses or that are at full rollout, to help you clean up flag code. | clientName |
Rules
| Tool | Description | Key parameters |
|---|---|---|
addOverrideRule | Adds an override rule that forces a variant for one entity, such as a user, visitor, or organization. | flagName, entity, entityValue, variantName |
updateOverrideRule | Changes the enabled state or assigned variant of an override rule. | ruleName, enabled, variantName |
addTargetingRule | Adds a conditional targeting rule that splits traffic across variants. Variant percentages must sum to 100. For version attributes, use versionValue in the payload instead of stringValue. | flagName, variantAllocations, payload, targetingKey, rolloutPercentage |
batchAddTargetingRules | Adds targeting rules to many flags in one call, using the same format as addTargetingRule. Returns succeeded and failed counts. | rules, completionLabels |
updateTargetingRule | Changes the enabled state, variant split, targeting conditions, or targeting key of a targeting rule. | ruleName, enabled, variantAllocations, targetingPayload, targetingKey |
deleteRule | Deletes a targeting or override rule. Entities that matched it fall through to the next rule. You can’t undo this. | ruleName |
getFlag with summary=false to find rule names.
Resolve
| Tool | Description | Key parameters |
|---|---|---|
resolveFlag | Resolves a flag for a client and entity and explains each evaluation step. Pass context to add attributes, such as country or plan, that targeting rules depend on. | flagName, clientName, entity, entityValue, context |
Evaluation Context
| Tool | Description | Key parameters |
|---|---|---|
getContextSchema | Returns a client’s evaluation context schema: every field you can target on, with its type. | clientName |
addContextField | Adds an entity field (used for bucketing) or attribute field (used for targeting conditions) to the evaluation context schema. | fieldName, fieldType, displayName, isEntity, entityReference, clientNames |
Identities and Users
| Tool | Description | Key parameters |
|---|---|---|
getIdentityInfo | Without arguments, returns the identity of the authenticated user. With identityName, returns details for that exact identity. | identityName |
searchIdentities | Searches users and groups by name and returns identities you can use as flag owners. Set kind to USER or GROUP, or omit it to search both. | query, kind, pageToken |
inviteUser | Invites a user to the Confidence account by email. | email, disableInvitationEmail |
getIdentityInfo or searchIdentities. Never build them from a display name or ID.
Metrics
| Tool | Description | Key parameters |
|---|---|---|
listMetrics | Lists metrics with their names, display names, entities, fact tables, and type specifications. | pageToken |
getMetric | Returns a metric. | metricName |
createMetric | Creates a metric that aggregates measurements from a fact table for an entity. | displayName, entity, factTable, typeSpec, description, metricId, preferredDirection |
createMetricCalculation | Starts an asynchronous calculation of a metric over a historical time window in the warehouse. | metric, exposureTable, startTime, endTime, metricCalculationId, labels |
getMetricCalculation | Returns the state of a metric calculation: STATE_RUNNING, STATE_COMPLETED, or STATE_FAILED. | name |
queryMetricCalculation | Returns per-dimension rows with count, mean, variance, and sum for a completed calculation. For a time series, set groupByDimensions and accumulateDimension to cohort_date. | name, groupByDimensions, accumulateDimension, timeZone |
listEntities | Lists entities: classes of instances you can identify and measure, such as users or transactions. | pageToken |
listFactTables | Lists fact tables with their entity columns, timestamp columns, measures, and dimensions. | pageToken |
listExposureTables | Lists exposure tables. Pass an exposure table name as exposureTable to createMetricCalculation. | pageToken |
Warehouses
| Tool | Description | Key parameters |
|---|---|---|
checkWarehouseExists | Checks whether the account has a data warehouse. Metric calculations require one. | None |
validateWarehouseConfig | Checks that access to datasets, schemas, and roles is set up before you create a warehouse. | warehouseType, configJson |
createWarehouse | Connects a BigQuery, Snowflake, Databricks, or Redshift warehouse. | warehouseType, configJson, warehouseId |
createCryptoKey | Creates a key for Snowflake key-pair authentication. You register the public key in Snowflake. | cryptoKeyId, kind |
createAssignmentTable | Creates an assignment table that defines how to read flag and experiment assignments from the warehouse. | displayName, sql, entityColumn, timestampColumn, exposureKeyColumn, variantKeyColumn, description, assignmentTableId |
createFlagAppliedConnection | Creates a connection that exports flag resolve events to the warehouse, so Confidence can write assignment data. | warehouseType, configJson, connectionId |
createEventConnection | Creates a connection that exports custom events to the warehouse, so Confidence can write event data for metrics. | warehouseType, configJson, connectionId |
Events
| Tool | Description | Key parameters |
|---|---|---|
createEventDefinition | Registers a typed schema for custom events that the SDK track() method sends. Confidence creates a matching fact table when events start arriving. | eventDefinitionId, schema, owner |
getEventDefinition | Returns an event definition with its schema, publish count, validation failures, and timestamps. | name |
listEventDefinitions | Lists event definitions with their schema fields and publish counts. | pageToken |
updateEventDefinition | Adds fields to an event definition schema. You can’t change the type of an existing field. | name, schema |
deleteEventDefinition | Deletes an event definition. Confidence rejects later events sent to it and doesn’t remove the warehouse fact table. | name |
queryEventsUsage | Returns hourly publish and validation failure counts for the last 1 to 7 days. Use it to verify that events arrive. | eventDefinitionName, daysBack |
Session Recordings
| Tool | Description | Key parameters |
|---|---|---|
listRecordingPolicies | Lists session recording policies with their resource names, display names, and clients. | pageSize, pageToken |
getRecordingPolicy | Returns a policy and its rules in evaluation order. | recordingPolicy |
createRecordingPolicy | Creates an empty recording policy for one client. Add rules separately. | displayName, clientName, aiAnalysisEnabled |
addRecordingRule | Adds an ordered rule to a policy, with targeting, a stable audience percentage, environments, and a per-session sample rate. Rules are drafts by default. | recordingPolicy, displayName, targetingKeySelector, targetingJson, stableAudiencePercentage, sessionSampleRate, environments, enabled |
setRecordingRuleEnabled | Enables or disables a recording rule. | recordingRule, enabled |