DEV Community: Obot AI

Free Live Workshop: Getting Started with Obot & MCP — From Zero to Production

Obot AI — Thu, 09 Jul 2026 10:30:00 +0000

From the Obot Academy: If you’ve been experimenting with the Model Context Protocol (MCP) but aren’t sure how to take it beyond a local demo, we’re hosting a free hands-on workshop July 29 @ 10am US PT that walks through the entire journey.

Register Now

What we’ll cover:

✅ What MCP is (and why it’s becoming the standard for AI agent integrations)
✅ Installing and configuring the Obot MCP Gateway
✅ Connecting AI agents to tools, APIs, and data sources
✅ Authentication, OAuth, and access policies
✅ Monitoring and operating MCP in production
✅ Live Q&A with Obot expert Chris Urwin

This is an interactive session — we’ll be building and demonstrating everything live.

Who it’s for

AI/ML Engineers
Backend & Platform Engineers
DevOps / DevSecOps
Technical Architects
Anyone building agents with or evaluating MCP for enterprise use

The Details:

📅 Date: Wed. July 29, 2026
🕐 Time: 10am PT / 1pm ET
🎟️ Click here to register

If you’re trying to figure out how to move from “I have an MCP server running” to “I can actually operate this in production,” this session is designed for you. We hope you'll join us!

Obot is an open-source MCP platform for hosting, securing, and managing MCP servers and AI agents at scale.

Best MCP Gateways for Engineering Teams (2026)

Obot AI — Fri, 03 Jul 2026 10:30:00 +0000

From the Obot AI team

Every engineering team that wires AI agents into production using MCP eventually hits the same inflection point: the setup that worked for two agents and three tools doesn't hold up across twelve agents, nine MCP servers, and a security review.

Credentials are inline. Access policies are per-server and inconsistently applied. The audit trail doesn't exist. IT wants to know what the agents are doing. Nobody has a clean answer.

That's where MCP infrastructure stops being optional — and where choosing the wrong gateway means rebuilding infrastructure that was supposed to be durable.

The MCP gateway landscape in 2026 is genuinely hard to navigate: purpose-built enterprise governance platforms, high-performance open-source proxies, managed cloud services, and legacy API gateway vendors who retrofitted MCP support into existing products. They are not interchangeable.

This guide covers what actually differentiates the leading options — with a feature comparison table, pricing overview, and a framework for making the right call based on your team's security, compliance, and operational requirements.

📓 Full guide · ⏱ 23-minute read

Originally published on Obot AI

Enterprise-Managed Authorization: How MCP Is Growing Up

Obot AI — Tue, 30 Jun 2026 20:02:37 +0000

By Bill Maxwell, Obot AI

MCP grew fast because it was easy to stand up and see immediate results. That ease of use is exactly what enterprises struggle with at scale.

OAuth — MCP's recommended authorization mechanism — works well for individual users granting consent to their own tools. It breaks down when an enterprise needs to manage access centrally: provisioning thousands of users, enforcing policy across departments, revoking access instantly when someone leaves.

Enterprise-Managed Authorization (EMA) is the MCP spec's answer to that problem. Now stable, EMA shifts control from individual OAuth consent flows to centralized, admin-managed authorization — the model enterprises already use for every other critical system.

This post covers how EMA works, what it still leaves to the gateway and governance layer, and where tools like Obot fit into the picture.

📓 Full article

Originally published on Obot AI

AI Governance Trends 2026: What Enterprise Teams Need to Know

Obot AI — Tue, 23 Jun 2026 15:38:09 +0000

By Shannon Williams, President & Co-founder of Obot AI

56% of CEOs say AI has delivered no measurable cost or revenue benefit. Only 1 in 8 sees improvement on both dimensions.

That's not a technology problem. It's a governance problem. AI is being deployed faster than the controls to manage it — and the gap between deployment speed and governance readiness is where the real risk lives.

Over the past several months the Obot team has had countless conversations with enterprise leaders across industries, and the same themes keep surfacing: shadow AI running unchecked on employee machines, agentic workflows that nobody has full visibility into, boards asking hard questions that "we have a policy document" can no longer answer.

This post maps where AI governance actually stands heading into the second half of 2026 — the regulatory shifts, the operational gaps, and what teams that are getting it right are doing differently.

Originally published on Obot AI

Obot Platform v0.23.0: LLM Gateway, MCP Search & Enterprise Upgrades

Obot AI — Fri, 19 Jun 2026 16:13:01 +0000

By Craig Jellick, Obot AI

Obot v0.23.0 just shipped, and it's one of our bigger releases.

The headline feature is the LLM Gateway — a centralized proxy that sits between your coding agents (Cursor, Claude Code, Windsurf, Copilot) and your LLM providers. One place to manage model access, enforce usage policies, and get visibility into what your agents are actually calling. If you're managing AI tooling across a team, this is the missing control layer.

But there's a lot more in this release:

→ MCP server search — find and connect MCP servers directly from within Obot without leaving the platform
→ GitOps for multi-user MCP servers — manage server configurations through Git workflows your team already uses
→ Enterprise licensing — new licensing controls for organizations running Obot at scale
→ Reworked UI — a cleaner, faster interface across the platform

Full breakdown in the release notes below.

📓 Full release notes · 🐙 GitHub release

Originally published on Obot AI

The Client Zoo Problem: Why Enterprise AI Needs Central Skills Management

Obot AI — Thu, 11 Jun 2026 19:53:14 +0000

By Bill Maxwell, Solutions Architect - Obot AI

Your support team built a great "triage a ticket" skill for Claude Code. It took weeks to get right. Meanwhile your sales engineers — working in Cursor — built the same skill from scratch last month. Theirs isn't as good, and neither team knows the other's exists.

This is where a lot of enterprise AI adoption has quietly landed. Multiple AI clients, duplicated workflows, no shared knowledge, no governance. The problem isn't that teams are using different tools — it's that reusable enterprise knowledge keeps getting trapped inside them.

This post breaks down why central skills management is the missing piece, and what it actually looks like to govern and distribute AI skills across an organization.

Originally published on Obot AI

How to Build and Publish a Skill Your Whole Team Can Use

Obot AI — Fri, 05 Jun 2026 13:42:00 +0000

By Craig Jellick, Obot AI

MCP servers get a lot of attention, but skills are where a lot of the day-to-day value lives for teams using AI agents. The problem is that most teams build skills in silos — one developer writes something useful, it lives on their machine, and nobody else benefits from it.

This tutorial walks through how to build a skill, publish it to a centrally managed catalog with Git-based versioning and access controls, and make it installable across any AI client your team uses — Cursor, Claude Desktop, Copilot, and more.

If you've been building with Obot or just exploring what centrally managed AI skills actually look like in practice, this is a good hands-on starting point.

📓 Full tutorial · 🐙 GitHub

Originally published on Obot AI

Who's Managing Your Team's AI Client Sprawl?

Obot AI — Wed, 03 Jun 2026 17:04:37 +0000

Something we keep running into in conversations with engineering and security teams: nobody really knows what AI clients and MCP servers are running on their developers' machines.

It's not negligence — it happened fast. Six months ago it was one or two tools. Now it's Cursor, Claude Desktop, Copilot, Windsurf, and a handful of MCP servers someone installed from a GitHub README. Each with its own config, its own credentials, nobody with a full picture.

We've been thinking hard about this problem at Obot AI and it's part of what drove our latest release. But we're curious where others are at — is fleet-wide visibility into AI clients and MCP servers on your radar yet, or is your team still figuring out more basic AI tooling questions first?

Obot Platform v0.22.0: Centrally Managed Skills, Fleet Scanning & Enterprise Controls

Obot AI — Wed, 03 Jun 2026 15:18:00 +0000

From the Obot AI team

We've been building Obot as an MCP Gateway from the start. v0.22.0 is where we go bigger.

The problem we kept hearing from teams: the "client zoo." Developers spread across Cursor, Claude Desktop, Copilot, Windsurf — each with its own config, its own local MCP servers, its own skill definitions. No standardization. No visibility for admins. Skills getting duplicated and inconsistently authored across every team.

v0.22.0, which shipped last week, tackles that directly with three new capabilities:

→ Centrally managed skills — admins publish a curated catalog; users install to any local AI client with a single command
→ Fleet scanning — obot scan gives you a live inventory of every AI client, MCP server, and skill running across your org
→ Enterprise MCP controls — managed image pull secrets, external secret integration, OAuth inspector, user-defined headers

This is the beginning of a broader vision: a control plane for your entire AI tooling stack, not just MCP.

📓 Full release notes · 🐙 GitHub release

Dangerous MCP OAuth Shortcuts are Ruining Security

Obot AI — Mon, 01 Jun 2026 16:31:52 +0000

By Adrian Goins, Obot AI

757 MCP servers compromised. 36% scored failing grades. Zero earned an A.

Those aren't projections — that's what a recent audit of real-world MCP OAuth implementations found. And the culprit isn't sophisticated attacks. It's shortcuts: client secrets hardcoded in frontend code, redirect URIs left wide open, PKCE skipped because it seemed like overkill.

OAuth is well understood at this point. So why are MCP implementations getting it so consistently wrong? Adrian Goins breaks down the specific patterns that are turning OAuth from a security control into a liability — and what a secure MCP OAuth implementation actually looks like.

Originally published on Obot AI.

Understanding the Model Context Protocol: A Beginner's Guide

Obot AI — Sat, 30 May 2026 14:08:00 +0000

By Stein Ove Helset, Obot AI

MCP (Model Context Protocol) keeps coming up everywhere in AI right now — but a lot of the explainers out there assume you already know what it is.

This guide starts from scratch. What problem did MCP actually solve? Why did AI integrations become such a fragmented mess before it existed? And how does the protocol work under the hood?

If you've been nodding along in conversations about MCP without fully following along — this one's for you.

Originally published on Obot AI

MCP Security Has Gone Mainstream

Obot AI — Fri, 29 May 2026 15:48:13 +0000

By Shannon Williams, President & Co-founder of Obot AI

A few days ago our outside counsel forwarded me a risk explainer on MCP from a major law firm's technology transactions group. It was the third in a series they've been running on MCP — laying out a NIST-based control framework for deploying MCP connectors in regulated enterprises.

That's not a developer audience. That's general counsel, CISOs, and procurement teams.

A year ago, MCP was something you explained to enterprise architects from scratch. Now the questions have shifted from "what is this" to "we already have this in production, our auditors are asking, what's the control framework."

MCP security has gone mainstream — and in this post I break down what that shift actually means and the four operational controls every organization deploying MCP needs to have in place.

Originally published on obot.ai