Skip to content

Blog

Practical guides on AI agent governance, the Agentic Gateway, and building production-ready agent infrastructure.

DeveloperStrategyArchitectureSecurityGovernanceFinOpsGetting Started
FeaturedDeveloper

5-Minute LangChain Observability: Cost, Latency, and Policy Without Rewriting a Line

Your LangChain agent is a black box in production. Here's how to get cost per run, full trace history, and policy enforcement by changing one environment variable: no SDK swap, no code rewrite.

Apr 28, 20265 min readGil Kal
FeaturedStrategy

Why CrewAI Agents Need an External Control Plane

CrewAI builds the crew. It does not answer for the fleet. What breaks past the first crew, what connecting actually involves, and when you genuinely do not need this.

Apr 25, 20264 min readGil Kal
FeaturedSecurity

Prompt Injection: Where Pattern Matching Fails

Pattern matching catches the scripted attacks and misses the novel ones. What runs before the model, and how to answer the question in a vendor risk pack.

Apr 20, 20265 min readGil Kal
FeaturedDeveloper

Connect Google ADK Agents to Dobby: MCP, A2A, and Webhook Integration (3 Paths)

Integrate agents built with Google's Agent Development Kit (ADK) via MCP, A2A, or webhooks. Monitoring, approval gates, and per-agent cost tracking: with working code examples.

Apr 15, 20268 min readGil Kal
FeaturedFinOps

Per-Agent LLM Cost Attribution: Who Spent What

Aggregate model spend tells you the bill, not the cause. How per-agent, per-team attribution works, and what a reviewer asks about it in a vendor review.

Apr 3, 20264 min readGil Kal
FeaturedArchitecture

Managing CrewAI, n8n and Custom Agents in One Place

Scheduling, webhook triggers, approval gates and audit trails for agents that already run on CrewAI, n8n, Make or your own Python. No agent code changes.

Apr 1, 20267 min readGil Kal
FeaturedStrategy

AI Agent Control Plane: The Layer That Has to Answer for the Fleet

Kubernetes gave containers a control plane. Agents need the equivalent, but the job is answering for what the fleet did, to someone outside the company. What that layer owes, and what to ask when evaluating one.

Mar 28, 20265 min readGil Kal
FeaturedFinOps

Cost Control for AI Agents: Three Layers and What Each One Misses

Budgets, provider quotas and model restrictions each stop a different failure, and none of them stops the first overage. What the three layers do, where they end, and what a vendor-risk reviewer asks about them.

Mar 22, 20265 min readGil Kal
FeaturedGetting Started

Connect Your First AI Agent to Dobby

Create a workspace, pick a region, connect an agent, and see its runs land. What the connection actually involves, and what you get once it does.

Mar 20, 20265 min readGil Kal
FeaturedGovernance

AI Agent Governance: The Four Controls a Vendor Review Asks You to Evidence

Policy, a halt, an audit trail and approval gates. What each control is, what record it has to leave behind, and why the record is the part that fails a review.

Mar 18, 20266 min readGil Kal
Architecture

Per-Tenant Gateway Profiles for Multi-Tenant AI Policy

One gateway, 50 tenants with different policies? Per-tenant gateway profiles give each tenant its own budget, models, and DLP via a 5-layer merge.

Apr 245 min read
Security

Data Handling: What Leaves, Where It Goes, Who Reads It

The data-handling cluster of a vendor questionnaire is answered by configuration a reviewer can check, not by a promise. Two scan layers, three actions, one invariant.

Apr 238 min read
Architecture

Change Management for AI: Proving What Was Live

A reviewer will ask which version served a decision last quarter. Prompt changes ship outside the release process, which is why that question is hard to answer.

Apr 227 min read
Governance

The Halt: Article 14's Hardest Requirement to Build

EU AI Act Article 14 requires the ability to interrupt or stop the system. Three scopes, why the read must fail closed, and what a halt has to record to count.

Apr 217 min read
Governance

Non-Human Identity: The Access Review Nobody Runs

Quarterly access reviews enumerate people. An autonomous agent holds delegated permissions and appears on no list: which is the finding, not the footnote.

Mar 288 min read
Governance

EU AI Act Article 14: Five Oversight Patterns

Article 14 asks for oversight a person can actually exercise. Five patterns that satisfy it, and what each leaves behind when a reviewer asks for proof.

Mar 265 min read
Architecture

Manage CrewAI, LangChain, and OpenAI Agents from One Dashboard

Your team uses CrewAI for orchestration, LangChain for RAG, and OpenAI Assistants for customer flows. Unified management, monitoring, and cost tracking: from one dashboard, across the frameworks your team already uses.

Mar 246 min read
Developer

MCP Protocol Explained: How AI Agents Call Real Tools and APIs (JSON-RPC)

The Model Context Protocol (MCP) gives AI agents structured access to tools and APIs via JSON-RPC. What it is, how it works, and why Anthropic, Claude, Cursor, and ChatGPT all speak it.

Mar 198 min read
Architecture

Agentic Gateway Explained: Auth, Cost Tracking, and Policy for Every LLM Call

The agentic gateway is a unified proxy that authenticates, meters, and enforces governance on every LLM and MCP request. Why every AI platform needs one, and how to build it.

Mar 158 min read

Ready to manage your AI agents?

Start free — no credit card required.

Get Started
AI Agent Governance & Compliance Blog | Dobby AI