WSJ: Nvidia launches OpenShell Reference Architecture to keep AI agents under control

Neo4j documents a property-graph database and its tooling at neo4j.com/docs/getting-started. The getting-started guide describes Cypher as Neo4j’s graph query language and links to data-modeling, import, application-integration, and visualization docs. It also points to Neo4j Graph Data Science, a library of graph algorithms, and GraphAcademy courses that include an Embedding & Vector Indexes tutorial. Neo4j MCP documentation lists running database options as Aura, Neo4j Desktop, or self-managed instances, and states that the official MCP server requires the APOC plugin on the instance it connects to. Use Neo4j when relationships among entities are the primary query pattern, and pair it with the neo4j-mcp catalog entry when agents need schema and Cypher access.

Claude is Anthropic's AI assistant based on the Constitutional AI and RLHF-aligned methodology. Launched in 2023, Claude emphasizes helpful, harmless, and honest interactions. It supports extremely long context windows of up to 200K tokens, making it effective for analyzing lengthy documents. Claude 3.5 Sonnet represents the mid-tier model with strong coding and reasoning capabilities. The iOS app and web interface provide easy access across devices.

Langfuse is an open-source product for LLM application observability: it ingests traces and spans from your stack, supports datasets and prompt/version workflows, and offers optional Langfuse Cloud or self-hosted deployment. It integrates with popular Python/JS SDKs and frameworks that emit OpenTelemetry-compatible telemetry, so teams can debug agent loops, compare prompt iterations, and monitor production quality metrics without building a custom analytics pipeline from scratch.

Neo4j documents an official Model Context Protocol server at neo4j.com/docs/mcp/current and in the neo4j/mcp repository. It gives MCP clients structured access to a Neo4j graph: get-schema introspects labels, relationship types, and property keys; read-cypher runs read-only Cypher checked with EXPLAIN and Neo4j query-type classification; write-cypher runs writes unless NEO4J_MCP_READ_ONLY=true; list-gds-procedures lists Graph Data Science procedures when GDS is installed. Prerequisites are a running Aura, Neo4j Desktop, or self-managed database plus the APOC plugin. Local clients use stdio via uvx neo4j-mcp-server or the Homebrew/binary neo4j-mcp, with NEO4J_MCP_URI, NEO4J_MCP_USERNAME, NEO4J_MCP_PASSWORD, and NEO4J_MCP_DATABASE. Aura uses HTTP at https://<instance-id>.mcp-instances.neo4j.io after Tool authentication is enabled in the Aura console. Docs also describe an HTTP transport mode that takes URI and credentials per request and selects the database from the /db/{databaseName}/mcp path.

Connects AI coding agents to Semgrep static analysis rules and findings so security reviews can start from actual scan results. Agents can look up rule definitions, triage findings, and verify fixes without pasting SARIF outputs. Integrates with CI/CD pipelines.

Structures The Wall Street Journal’s October 21, 2026 live-coverage card on Nvidia’s OpenShell Reference Architecture into an agent-governance checklist. The workflow keeps only what reporter Mauro Orru’s item states: Nvidia is launching a system it says will help enterprises keep AI agents in check after a string of incidents in which the technology went rogue; the system is dubbed Nvidia OpenShell Reference Architecture; it brings isolation and security capabilities to build AI agents and observe how they behave; CEO Jensen Huang said that as enterprises prepare to deploy billions of AI agents, safety and control must be built into the computing architecture. The Journal card was last updated October 21, 2026 at 8:49 p.m. ET. Steps require confirming any further product claims—hardware names, latency figures, partner lists—against the Journal text you can access and against NVIDIA’s own materials, instead of treating vendor numbers as WSJ findings.