{"resourceId":"cisco-iq-rest-apis-mcp-server-agent-benchmark-20261104","versions":[{"version":"external-78c98eadc5ab5f0f8d4c8c4ebcc7531f88c085fd94e511ca1a40e876f6671818","resource":{"id":"cisco-iq-rest-apis-mcp-server-agent-benchmark-20261104","title":"Cisco says Cisco IQ now exposes production REST APIs and an MCP server, and that scripted tests cut agent tool calls from 256 to 36","organization":"Cisco","sector":"Vendor; IT operations and lifecycle intelligence relevant to SLED IT","geography":"Global","publishedAt":"Cisco Blogs, Nov. 4, 2026","publicationDate":"2026-11-04","eventDate":null,"sourceName":"Cisco Blogs","sourceLabel":"Cisco IQ: Engineering APIs for the Agentic Era","sourceUrl":"https://blogs.cisco.com/customerexperience/cisco-iq-engineering-apis-for-the-agentic-era","evidenceClass":"vendor-claim","outcomeClass":"emerging","topics":["developers-agents","infrastructure","data-security"],"finding":"Cisco says Cisco IQ, generally available since April 2026 and deployed by over 16,000 customers, now offers production REST APIs and a Model Context Protocol server built from one canonical domain model. Agents connecting over MCP inherit the calling principal's permissions through Cisco Unified Identity, with action-level permissions and validated queries that fail loudly rather than being rewritten. In scripted tests on a 17-scenario benchmark corpus, Cisco reports tool calls fell from 256 to 36 and approval prompts from 181 to 1. Live-agent results are not yet published.","sledRelevance":"Relevant to state, campus and district IT teams that use Cisco IQ for asset, advisory and contract lifecycle data and want AI assistants to query it without creating a separate permission model.","evidence":"Vendor-authored engineering blog. The benchmark is Cisco's own scripted test on 17 scenarios with no published method, dataset or independent replication; Cisco says live-agent data is still being measured.","architectureImplications":"Agents query lifecycle, advisory and contract data through MCP using the same domain model and RBAC as REST clients.","governanceImplications":"Agent access inherits existing human role permissions, so role design and least-privilege review become the main control point.","securityPrivacyImplications":"Agents receive only what the calling principal can see; agencies must still check which principal agents run under and log their queries.","caveats":"Self-reported scripted benchmark; 16,000-customer figure unverified; no SLED examples.","streamIds":["cisco"],"roles":{"sales":"Sales framing: position this as governed data access for AI assistants, not proven productivity. Cisco says agents reach Cisco IQ through the same permissions as people and reports fewer tool calls in its own scripted tests. Ask SLED IT leaders which lifecycle and security-advisory questions consume staff time, and propose a scoped read-only pilot. Do not quote the 256-to-36 figure as customer results; it is a vendor lab benchmark.","engineering":"Engineering framing: Cisco IQ exposes OpenAPI 3.2 REST endpoints and an MCP server generated from one domain model, with action-level permissions, strict query validation, RFC 9457 errors, ETags and partial-success bulk writes. In a pilot, connect an agent under a read-only service principal, confirm it cannot exceed that role, log every MCP call, and run your own question set to compare call counts and accuracy with the vendor benchmark.","delivery":"Delivery framing: begin with read-only lifecycle and advisory queries for one team, such as end-of-life and security-advisory exposure reports. Define which principal the agent uses, who approves any write scope, and how queries are logged. The IT asset manager owns the pilot with security review. Track answer accuracy and staff time saved against a manual baseline before granting write permissions."},"roleDetails":{"sales":{"customerProblem":"IT teams spend time stitching asset, advisory and contract data to answer lifecycle and exposure questions.","stakeholders":["IT director","IT asset manager","Security officer"],"discoveryQuestions":["Do you use Cisco IQ today?","Which lifecycle or advisory questions take the most staff time?","Under what identity would an AI assistant run?"],"valueHypothesis":"Governed agent access could speed routine lifecycle answers; only vendor scripted data exists.","potentialEngagement":"A read-only Cisco IQ MCP pilot with a local question set.","unsupportedClaims":["Customer productivity gains","Live-agent performance","Independent benchmark replication"]},"engineering":{"fit":"Fits Cisco IQ customers wanting agent access under existing RBAC.","architectureIntegration":"REST and MCP surfaces from one canonical model; Cisco Unified Identity authorization.","prerequisites":["Cisco IQ subscription","Cisco Unified Identity","Agent platform with MCP support"],"constraints":["Benchmark is vendor scripted","Write operations need careful scoping"],"security":"Agents inherit principal permissions; use dedicated least-privilege principals and log calls.","proofOfValue":["Run local question set and record call counts","Verify role boundary enforcement","Compare answers to manual reports"]},"delivery":{"work":"Read-only agent pilot over Cisco IQ data.","dependencies":["Entitlement confirmation","Identity setup","MCP-capable agent"],"ownership":"IT asset manager with security review","skillsAdoption":"Staff need to learn to validate agent answers.","governanceCheckpoints":["Principal and scope approval","Query logging review","Write-scope approval gate"],"acceptanceCriteria":["Proposed: agent answers match manual reports on test set","Proposed: no access beyond assigned role in tests"],"risks":["Over-trust in vendor benchmark","Excessive permissions","Inaccurate answers"]}},"retrievedAt":"2026-11-05T04:09:01Z","enrichedAt":"2026-11-05T04:10:39Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"May let smaller teams answer lifecycle and advisory questions faster; no workforce evidence.","procurementImplications":"Check whether API and MCP access is included in existing Cisco IQ entitlements and how partner managed services use the data.","operatingModelImplications":"Encourages treating AI assistants as API clients under existing identity governance rather than special integrations.","sourceVerification":{"openedUrl":"https://blogs.cisco.com/customerexperience/cisco-iq-engineering-apis-for-the-agentic-era","referenceExcerpt":"In scripted tests across a benchmark corpus of 17 complex enterprise scenarios, this architecture delivered decisive operations improvements:","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}