February 27, 2026

MCP Weekly: Open-Standard Momentum, Agent-Assignable Jira, and Anthropic’s Defense Line

Conformance, connectors, and real-world constraints

In this blog, we look at major MCP developments from February 20th to February 27th, 2026, This issue covers open standard governance, platform launches from Atlassian and Anthropic, a confrontation between Anthropic and the U.S. Department of War, Claude product updates, and the latest enterprise tooling and funding news.

TL;DR

The Agentic AI Foundation welcomed 97 new members, bringing its total to 146, with the MCP Dev Summit confirmed for April 2-3 in New York. Atlassian made AI agents first-class members in Jira and expanded Rovo to connect to third-party tools via MCP. Anthropic CEO Dario Amodei publicly refused two demands from the U.S. Department of War, mass domestic surveillance and fully autonomous weapons, with the department threatening to invoke the Defense Production Act.

Anthropic acquired Vercept to advance computer use and expanded Cowork with finance plugins including live connections to FactSet and MSCI. Claude Code gained full lifecycle automation from preview through to pull request merge. Figma launched an MCP server for Codex, bridging design and live code. ServiceNow launched autonomous AI specialists with Claude as the default model. Basis raised $100M at a $1.15B valuation for AI-powered accounting.

Major Updates of the Week

AAIF Grows to 146 Members, MCP Dev Summit Set for April

The Agentic AI Foundation added 97 new organizations, including JPMorgan Chase, American Express, Red Hat, Equinix, and Akamai, bringing total membership to 146. David Nalley from AWS has been named Governing Board Chair. The foundation governs MCP, goose, and AGENTS.md. The MCP Dev Summit is confirmed for April 2-3 in New York City, with over 95 sessions covering security, conformance, and enterprise-scale design.

Atlassian Makes Jira and Rovo Agent-Native via MCP

Jira now treats AI agents as first-class team members , assignable to tickets, reachable via comments, and triggerable at workflow transitions, with all activity logged for compliance. Rovo connects to Figma, Amplitude, Box, Canva, and GitHub via MCP with no custom code. Admins control which skills are active, and teams can build custom agents in Rovo Studio by combining tools from across their stack.

Anthropic Updates

Vendor / Product	Key Action / Feature	Significance
Anthropic and U.S. Department of War	CEO Dario Amodei confirmed Claude is deployed for defense and intelligence work but refused to support mass domestic surveillance or fully autonomous weapons. The Department of War threatened to invoke the Defense Production Act.	Sets a public precedent for how frontier AI labs handle defense contracts and hard safety limits.
Anthropic acquires Vercept	Acquired a startup focused on AI perception and software interaction. Claude Sonnet 4.6 already reaches 72.5% on OSWorld.	Moves Claude toward operating across desktop apps and browser workflows without APIs.
Cowork finance plugins	Private plugin marketplace with MCP connectors to FactSet, MSCI, S&P Global, and LSEG. Pre-built templates cover investment banking, equity research, private equity, and wealth management.	Targets the data-to-deliverable pipeline in finance, with Claude running across Excel and PowerPoint in one session.
Responsible Scaling Policy 3.0	Separates internal safety commitments from industry recommendations, with a public Frontier Safety Roadmap and Risk Reports every three to six months.	Provides a more transparent, audit-ready framework for enterprise and government stakeholders.
Distillation attack disclosure	Three Chinese labs used 24,000 fraudulent accounts to extract 16 million training exchanges from Claude, targeting agentic reasoning and coding capabilities.	Signals tighter API access controls and behavioral fingerprinting across frontier model providers.

Claude Updates

Product	Key Action / Feature	Significance
Claude Code on desktop	Previews running web apps, reviews code diffs with inline comments, monitors GitHub pull request status, auto-fixes failed checks, and merges approved pull requests in the background. Context carries across desktop, command line, and mobile.	Moves Claude Code from a coding assistant to a full development lifecycle tool.
Claude Code for COBOL	Published guidance on automating legacy dependency mapping and documentation recovery for COBOL systems.	Addresses the phases that have historically made COBOL modernization too expensive to justify.

Figma and Codex: Design Meets Code via MCP

Figma launched a new MCP server for the Codex desktop app, creating a two-way bridge between design and live code. Developers can pull Figma layout data into Codex agents to generate production-ready code, and run the process in reverse, capturing a live web interface and converting it back into editable Figma layers in seconds. The MCP connection keeps both sides in sync as styles and variables are updated.

Other Updates

Vendor / Product	Key Action / Feature	Significance
ServiceNow	Launched autonomous AI specialists that own full job roles, with Claude as the default model for the ServiceNow Build Agent.	Resolves over 90% of internal IT requests autonomously, combining model reasoning with deterministic workflow rules. GA expected Q2 2026.
SambaNova and Intel	Launched the SN50 chip with a claimed 5x speed advantage over NVIDIA B200, backed by $350M Series E funding.	Targets production inference where cost-per-token and latency matter more than raw training scale.
Basis	Raised $100M Series B at a $1.15B valuation, backed by Accel and Google Ventures. Serves 30% of the top 25 U.S. accounting firms.	Runs long-horizon agents for tasks like partnership tax returns, addressing a structural talent shortage in accounting.
Google Opal	Added an agentic step that self-selects tools, retains memory across sessions, and asks clarifying questions before proceeding.	Moves Opal from fixed automation to adaptive, reasoning-based workflows.

My Thoughts: MCP Has Entered Its Production Era

AAIF hitting 146 members is the clearest proof yet that MCP has crossed into shared infrastructure. When JPMorgan, Amex, Red Hat, Equinix, and Akamai all show up under the same umbrella, the center of gravity shifts toward conformance, security, and operational discipline. That’s why the April 2-3 Dev Summit matters. This ecosystem is entering its “production era,” where test suites, SDK alignment, reference architectures, and governance decide what gets adopted enterprise-wide.

Atlassian’s Jira update is the most practical signal in this issue. “Agent-assignable” work changes behavior: agents stop living in side panels and start living inside the workflow itself, reachable in comments and triggerable at transitions. The part that will drive real adoption is the control plane—activity logs for auditability, admin-scoped skills, and an opinionated path to connecting third-party tools via MCP. That’s the bundle enterprises buy: velocity that still looks governable when compliance asks, “show me exactly what happened.”

The Anthropic and Department of War standoff is the most consequential agentic story of the week because it makes model policy operational. A provider’s red lines become your system’s edges, shaping what can be automated, how exceptions get handled, and where humans must stay in the loop. Buyers are going to start evaluating frontier models the way they evaluate platforms: clear constraints, enforceable controls, and predictable escalation paths. The vendors that win trust will be the ones who can prove those boundaries with telemetry and still keep workflows moving when the model draws a hard line.

‍

Om Shree

Technical Evangelist

About Om Shree

Om Shree is a researcher, technical writer, and AI evangelist who focuses on making complex AI and agent workflows easier to understand. Om's passion is breaking down emerging technologies into clear, practical insights. He's excited to provide useful in-depth research that supports product planning and helps developers navigate new tools and systems with ease.

Customized Plans for Real Enterprise Needs

Gentoro makes it easier to operationalize AI across your enterprise. Get in touch to explore deployment options, scale requirements, and the right pricing model for your team.

Get in Touch