generated: '2026-08-01' method: derived source: >- Derived by aligning Conviva's published MCP surface (mcp/conviva-mcp.yml, tool names taken verbatim from Conviva's own Agent Skills in skills/) with the documented REST surface at api.conviva.com (docs.conviva.ai API guides). purpose: >- Bind each agent-facing MCP capability to the REST capability that backs it, and record where the two surfaces DIVERGE. Conviva is an unusually clean example of MCP and REST being different projections of one data core rather than one wrapping the other: the REST surface is VSI/Insights shaped (metrics, sessions, alerts, filters, policy, privacy), while the DPI MCP surface is Context Center shaped (semantic knowledge search, asset hydration, predefined-metric execution, Nexa analysis, behavior segments). Neither is a superset. important_caveat: >- Conviva publishes NO OpenAPI, so there are no operationIds to bind to. Every `rest:` row below therefore names a documented REST PATH, not an operationId, and `binding: rest-path` marks that distinction. Nothing here is invented: every path appears in a Conviva API guide and every tool name appears in a Conviva-published Agent Skill. Because tools/list is auth-gated on both servers, no tool inputSchema is asserted anywhere. surfaces: rest_openapi: null # no OpenAPI published (Conviva says specs are "being wired up") rest_docs: https://docs.conviva.ai/connect-data/apis/ rest_base: https://api.conviva.com graphql: null # no GraphQL surface found mcp: - url: https://mcp.conviva.com/mcp version: 1.0.0 tools_list: gated (401) services: ["/vsi/metrics", "/vsi/ai-alerts", "/vsi/sessions", "/dpi/metrics", "/dpi/ai-alerts"] tool_names_published: false - url: https://dpi-mcp.conviva.com/mcp version: 1.1.0 tools_list: gated (401) services: ["/insights", "/nexa", "/context-center", "/session-replay", "/metric-query"] tool_names_published: true tool_name_source: https://github.com/Conviva/mcp-marketplace (Agent Skills) crosswalk: - tool: metric-query-run server: dpi-mcp category: metrics rest: ["/insights/3.0/metrics/{metric}", "/insights/3.0/metrics/custom-selection", "/insights/3.0/metrics/{metric}/group-by/{dimension}"] binding: rest-path confidence: medium note: >- Both run a metric over a time window with an optional dimensional breakdown, but the identifiers differ fundamentally. REST addresses a Conviva-DEFINED metric by NAME in the path and breaks down by a named dimension; the MCP tool addresses a CUSTOMER-DEFINED Context Center metric by opaque m_* id and breaks down by dimension_* ids. They are analogous capabilities over different catalogs, not the same operation. - tool: context-center-asset-list server: dpi-mcp category: metadata rest: ["/3.0/metrics/_meta/references/filters"] binding: rest-path confidence: low note: >- Both enumerate reference metadata, but REST exposes only filter references under _meta while the MCP tool pages through the full Context Center asset catalog (pattern, metric, dimension, segment, template, insights-finder). Weak, partial overlap only. - tool: insights-behavior-segment-list server: dpi-mcp category: insights rest: [] binding: none confidence: high note: Listed here for completeness; no REST equivalent — see mcp_only[]. mcp_only: - tool: context-center-nexa-semantic-search reason: >- Vector/ANN semantic search over an LLM-generated knowledge layer. There is no REST search endpoint of any kind on api.conviva.com — the REST surface has no discovery-by-meaning primitive. - tool: context-center-nexa-categories-list reason: Knowledge-layer category taxonomy; no REST counterpart. - tool: context-center-nexa-knowledge-get reason: Fetches an LLM-generated k_* knowledge summary; the knowledge layer is MCP-only. - tool: context-center-asset-get reason: >- Hydrates a customer-defined Context Center asset (pattern / metric / critical-event / dimension / segment / template) by id. The Context Center asset model is not exposed over REST at all. - tool: nexa-analyze reason: >- Async open-ended AI analysis ("why did X change"). A server-side composite with no REST equivalent; requires a fallbackReason enum. - tool: nexa-analyze-result reason: Retrieves an async Nexa job result; no REST job API exists. - tool: insights-behavior-segment-list reason: Insights behavior segments (formerly "patterns") are computed offline by Insights and are not published on the REST surface. - tool: insights-behavior-segment-get reason: Same — a stored Insights behavior segment with pre-computed insights, evidence and conversion barriers, unavailable over REST. - tool: identity-c3-account-list reason: >- Lists the c3 accounts an OAuth caller is entitled to. REST has no account-discovery endpoint because REST credentials are minted per c3 account in Pulse — the account is implicit in the key. rest_only: - capability: Viewer session detail rest: ["/insights/3.0/sessions/content", "/insights/3.0/sessions/ad", "/insights/3.0/sessions/content/viewer", "/insights/3.0/sessions/ad/viewer"] note: >- No session tool appears in any published skill on dpi-mcp. mcp.conviva.com advertises a /vsi/sessions service, so a session tool very likely exists there — its tool names are simply not published anonymously. Recorded as rest_only with that caveat rather than guessed. - capability: Real-time metrics rest: ["/insights/3.0/real-time-metrics/{metric}"] note: 15-minute real-time window; no published MCP tool for it (again, /vsi/metrics may cover it). - capability: AI alerts rest: ["/insights/2.6/ai-alerts/content-metrics", "/insights/2.6/ai-alerts/ad-metrics", "/insights/2.6/ai-alerts/content-metrics/{alert_id}", "/insights/2.6/ai-alerts/ad-metrics/{alert_id}"] note: >- mcp.conviva.com advertises /vsi/ai-alerts and /dpi/ai-alerts services and the MCP docs say alerts are covered, but no alert tool name is published. - capability: Saved filter management rest: ["/bulk_filters"] note: Bulk create/read/update/delete of filters. No MCP tool — writes are absent from the published agent surface entirely. - capability: Precision policy control rest: ["/precision/v1.0/policies", "/precision/v1.0/policies/activate", "/precision/v1.0/policies/reorder"] note: Write/control-plane operations gated behind Precision Admin. No MCP exposure. - capability: PII opt-out / privacy requests rest: ["/pii-opt-out"] note: Data-subject deletion workflow. Deliberately absent from the agent surface. - capability: Integration validation timeline rest: ["/validation/v2/timeline"] note: QA/sensor-integration tooling. No MCP exposure. - capability: Session replay rest: [] note: >- Inverse case — /session-replay is advertised in the dpi-mcp service descriptor but has no REST endpoint AND no published tool name. Neither surface documents it publicly. divergence_summary: >- The two surfaces barely overlap. REST is the Conviva-defined VSI measurement and administration API; MCP is the customer-defined Context Center semantic layer plus AI analysis. Only metric-query-run has a genuine (and still imperfect) REST analogue. Every write operation Conviva exposes over REST — filters, Precision policy, PII opt-out — is absent from the agent surface, which is read-only by construction. That is a deliberate and defensible agent-safety posture, and worth recording as such rather than as a coverage gap. coverage: tools_named: 11 tools_bound_to_rest: 2 tools_partially_bound: 1 mcp_only: 9 rest_capabilities_total: 8 rest_capabilities_with_a_tool: 1 mcp_servers: 2 mcp_services_advertised: 10 mcp_services_with_published_tool_names: 5 rest_operations_total: unknown # no OpenAPI to count operations from agent_surface_is_read_only: true x-evidence: fetched: '2026-08-01' mcp_tools_list_http_status: 401 openapi_probe_result: no spec found on any host