DEV Community: Obot AI

Obot Platform v0.24.0: Obot Sentry, AI Audit Logs & MCP Improvements

Obot AI — Mon, 27 Jul 2026 12:39:00 +0000

By Craig Jellick, VP Engineering & Co-founder of Obot AI

v0.24.0 has shipped, and we're proud to highlight Obot Sentry — a lightweight companion agent that enrolls a workstation with Obot, runs scheduled device inventory scans, and installs managed audit hooks for Claude Code, Codex, VS Code, and Cursor.

Why Obot Sentry? Teams are trying to understand what AI is doing across their organizations, but activity is scattered. Some traffic goes through an LLM gateway. Some happens in local AI clients that never cross a centralized endpoint. Some is in MCP servers, some is direct model calls. No single capture point covers all of it.

v0.24.0 adds three pieces to address that:

→ Obot Sentry — device-level audit hooks for local AI clients; first MDM path is Microsoft Intune for Windows
→ Full LLM Gateway audit logs — every model request logged with user, provider, model, token usage, duration, and encrypted content; exportable to S3, GCS, or Azure
→ AWS Bedrock, Azure, and Generic Responses providers — broader model provider coverage without a different governance model for each

Plus: composite MCP servers through GitOps, a significantly smaller MCP proxy footprint, built-in local auth, and in-app getting started guides.

📓 Full release notes · 🐙 GitHub v0.24.0

Give Obot a try - no strings, no card needed, full enterprise edition, free for 2 weeks.

Originally published on Obot AI

MCP Is Growing Up: What the 2026 Roadmap Actually Changes

Obot AI — Fri, 24 Jul 2026 14:50:56 +0000

By Bill Maxwell, Obot AI

The MCP spec dropping July 28 is the largest revision since MCP launched — and as a team that's been building enterprise infrastructure on top of MCP since the beginning, we've been watching this roadmap closely.

The headline change is a stateless core. The initialize/initialized handshake is gone. Protocol metadata now travels with every request instead of being negotiated once at connection time. For teams running Obot, this is a direct win: no more sticky sessions, no shared session store, no extra routing complexity at the gateway layer.

But the stateless core is just one part of a larger shift. The 2026 roadmap also formalizes an extensions system, hardens OAuth/OIDC to match how it's actually deployed in production, and establishes a proper twelve-month deprecation lifecycle. MCP is replacing early protocol-specific machinery with proven infrastructure patterns.

What's equally telling is what the roadmap leaves deliberately undefined: audit trails, SSO-integrated authorization, gateway behavior, and configuration portability are explicitly called out as enterprise-readiness concerns — and described as an open invitation for the ecosystem to fill. That's exactly the space Obot is built for, and this release makes the foundation beneath us significantly stronger.

Obot Software Architect Bill Maxwell breaks down every major change and what it means for teams running MCP in production today.

📓 Full article

Originally published on Obot AI

What Is AI Governance? A Framework for Enterprise MCP Teams

Obot AI — Tue, 21 Jul 2026 16:10:08 +0000

From the Obot AI team

An organization can have its EU AI Act risk tiers mapped, ISO/IEC 42001 certification underway, and a model review board that meets every month — and still have no answer to a basic operational question: which of the MCP servers its agents can reach were ever vetted for anything?

The problem is structural. Most enterprise AI governance frameworks — the EU AI Act, NIST AI RMF, ISO/IEC 42001 — were built to govern models: their training data, outputs, and risk classification. None of them were designed to answer for an agent's runtime decision to call an MCP tool. The entire tool layer is effectively ungoverned even in organizations with mature AI governance programs.

This post maps exactly where general AI governance frameworks stop at the model boundary, and what an MCP-specific governance layer actually needs to include: policy, risk classification, registration-time vetting, runtime monitoring, audit trails, and a maturity model for managing the tool layer.

If your organization is running MCP in production and has a governance program that doesn't explicitly cover it — this is the framework you're missing.

⏱ 18-minute read · 📓 Full article

Originally published on Obot AI

[Boost]

Obot AI — Fri, 10 Jul 2026 12:58:45 +0000

Obot AI

Jul 9

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

#agents #ai #mcp #tutorial

1 min read

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?