agentops-cockpit 1.4.5

Production-grade Agent Operations (AgentOps) Platform

🕹️ AgentOps Cockpit

🌐 Official Website & Live Demo

"Infrastructure gives you the pipes. We give you the Intelligence."

The developer distribution for building, optimizing, and securing AI agents on Google Cloud.

---

📽️ The Mission

Most AI agent templates stop at a single Python file and an API key. The AgentOps Cockpit is for developers moving into production. It provides framework-agnostic governance, safety, and cost guardrails for the entire agentic ecosystem.

• Governance-as-Code: Audit your agent against Google Well-Architected best practices with the Evidence Bridge—real-time citations for architectural integrity.
• SME Persona Audits: Parallelized review of your codebase by automated Principal SMEs across FinOps, SecOps, Architecture, and Quality.
• Agentic Trinity: Dedicated layers for the Engine (Logic), Face (UX), and Cockpit (Ops).
• A2A Connectivity: Implements the Agent-to-Agent Transmission Standard for secure swarm orchestration.
• MCP Native: Registration as a Model Context Protocol server for 1P/2P/3P tool consumption.

---

🏗️ The Agentic Trinity

We divide the complexity of production agents into three focused pillars:

graph TD
   subgraph Trinity [The Agentic Trinity 2.0]
       E(The Engine: Reasoning)
       F(The Face: Interface)
       C(The Cockpit: Operations)
       S{Sovereignty & Compliance}
   end
   E <--> C
   F <--> C
   E <--> F
   E -.-> S
   F -.-> S
   C -.-> S
   style Trinity fill:#f8fafc,stroke:#334155,stroke-width:2px
   style S fill:#0ea5e9,color:#fff,stroke:#0284c7

• ⚙️ The Engine: The reasoning core. Built with ADK, FastAPI, and Vertex AI.
• 🎭 The Face: The user experience. Adaptive UI surfaces and GenUI standards via the A2UI spec.
• 🕹️ The Cockpit: The operational brain. Cost control, semantic caching, shadow routing, and adversarial audits.

---

🕹️ v1.4.4: The "Sovereign Evolution" Release (NEW)

Evolving into a full Lifecycle Management Platform for AI Agents. See the v1.4.4 Release Notes.

• 🛡️ Validation Automation: Integrated /validate workflow to enforce 100% SME approval and regression safety.
• 🧗 Autonomous Evolution (10X): The "PR Closer" mode. Surgically fixes detected gaps and creates a hardened deployment branch automatically.
• 🕵️ Shadow Mode (10X): Differential reasoning analysis to detect drift, latency, and cost delta between agent versions.
• 🧗 Atomic RAG & Hardening: Restricted retrieval context to top 5 results to maximize reasoning density and prevent token bloat.
• 🛠️ Poka-Yoke Tooling: Hardened tool schemas using Literal types to "mistake-proof" agentic tool calling.
• 💰 SME Consensus 2.0: Unified approval engine requiring a unanimous "Sovereign Standard" from all 11 Principal SMEs.

---

🚀 Key Innovation: The "Intelligence" Layer

🛡️ Red Team Auditor (Adversarial SRE)

Don't wait for your users to find prompt injections. Use the built-in Adversarial Evaluator to launch self-attacks against your agent, testing for PII leaks, instruction overrides, and multilingual jailbreaks.

🧠 Hive Mind (Semantic Caching)

Reduce LLM costs by up to 40%. The Hive Mind checks for semantically similar queries in 10ms, serving cached answers for common questions without calling the LLM.

🏛️ Arch Review & Autonomous Evolution

Every agent in the cockpit is graded against a framework-aware checklist. The Cockpit intelligently detects your stack and runs a tailored Architecture Review. v1.3 introduces Autonomous Evolution—the ability to synthesize code fixes directly from audit findings.

🕹️ MCP Connectivity Hub (Model Context Protocol)

Stop building one-off tool integrations. The Cockpit provides a unified hub for MCP Servers. Connect to 1P/2P/3P tools via the standardized Model Context Protocol for secure, audited tool execution. Start the server with make mcp-serve.

🗄️ Situational Database Audits

The Cockpit now performs platform-specific performance and security audits for AlloyDB, Pinecone, BigQuery, and Cloud SQL.

---

🛡️ Advanced Governance & Discovery (v1.3.5)

Modern agents don't just live in agent.py. The Cockpit uses a centralized Discovery Engine to intelligently map your project:

• .gitignore Compliance: Zero-noise scanning that respects your project's ignore rules.
• Multi-Target Logic: Define targets: [] in cockpit.yaml to audit distributed agents in a single pass.
• Template Isolation: Automatically ignores raw template placeholders (e.g., Jinja/Cookiecutter) to focus on the active implementation.
• Artifact Store: All data (SARIF, Evidence, HTML) is now sovereignly stored in the .cockpit/ directory.

---

⌨️ Master Command Registry

The Cockpit is available as a first-class CLI and a comprehensive Makefile-based operational toolkit.

Registry	Description
🕹️ Makefile Commands	Standard local development and orchestration shortcuts.
🚀 UVX Master Guide	Portable, zero-install commands for CI/CD and automation.

---

🧑‍💼 Principal SME Persona Approvals

The Cockpit now features a Multi-Persona Governance Board. Every audit result is framed through the lens of a Principal Engineer in that domain:

• 🏛️ Architecture
• 💰 FinOps
• 🛡️ Red Team
• 🧗 Quality
• 🌐 Infrastructure
• 🎭 UX/UI

---

🚀 Production Readiness Auditor

The Cockpit serves as the final gate before production deployment. make deploy-prod triggers a deep benchmark of the entire ecosystem:

1. v1.4.1 Deep System Audit: Benchmarks models (Gemini 3 Pro/Flash) and logic.
2. Stress Testing: Load testing endpoints to ensure concurrency safety.
3. Red Team Verification: Adversarial security scans for prompt injection and PII.
4. Resiliency Check: Verifies @retry logic and timeout guards are active.

---

🤝 Ecosystem & Attribution

The AgentOps Cockpit is designed to leverage and secure the best-of-breed tools in the Google Cloud ecosystem. We explicitly acknowledge and leverage the excellent work from:

• GoogleCloudPlatform/agent-starter-pack: We leverage this as a core reference for the Agent Development Kit (ADK) patterns and Vertex AI Agent Engine integration.
• A2UI Protocol: Standardized Generative UI handshake for building adaptive, agentic user interfaces.
• A2A Standard: Agent-to-Agent Transmission Protocol for secure swarm intelligence and inter-agent communication.
• Model Context Protocol (MCP): Our unified tool execution standard, enabling portable and secure 1P/2P/3P integrations.
• LangChain & LangGraph: Foundational libraries for stateful, multi-agent reasoning loops and graph-based orchestration.
• CrewAI: Multi-agent framework used as a reference for collaborative task execution and role-playing agents.
• Firebase: Provider for enterprise-grade hosting and global distribution of the Face layer.
• Google Cloud Run & GKE: High-scale orchestration platforms for the Engine and cluster-wide agent fleets.
• Vertex AI SDK: The backbone for frontier reasoning (Gemini 3) and enterprise-grade model governance.
• Tenacity: The gold-standard library for the exponential backoff and resiliency patterns we enforce.
• Rich: Modern visualization engine that powers the high-fidelity Cockpit CLI experience.

---

Reference: Google Cloud Architecture Center - Agentic AI Overview