February 20, 2026

MCP Weekly: Slack Goes Agent-Native, Google Database MCP, OpenAI Personal Agents

Security moves from a checkbox to the agenda, as Slack, Microsoft, and Google harden the agent stack

In this blog, we look at major MCP developments from February 13th to February 20th, 2026, including platform updates from Slack, Google and OpenAI, a major new Claude release from Anthropic, enterprise partnerships, and new startup funding in the MCP infrastructure space.

TL;DR

Slack launched new agent-native MCP servers and Real-time Search APIs, prioritizing granular security for AI access to workspace data. Google expanded Managed MCP support across its core database portfolio and released Gemini 3.1 Pro. Google also upgraded Gemini 3 Deep Think with world-class scientific and coding reasoning, making it available via an early access API.

Anthropic's Claude Sonnet 4.6 now delivers flagship-level performance at mid-tier pricing, powering major strategic partnerships with Infosys (enterprise agents) and the Government of Rwanda (national AI infrastructure). Enterprise security is tightening with Microsoft’s new Security Dashboard for AI. OpenAI hired Peter Steinberger, creator of open-source agent framework OpenClaw. Key infrastructure funding went to platforms like Manufact ($6.3M) and Autosana ($3.2M).

Major Updates of the Week

Slack Launches Agent-Native Infrastructure

Slack released two new developer tools this week: an MCP server built specifically for AI agents and a Real-time Search API that replaces its older Data Access API. The MCP server returns natural language responses designed for model consumption, while the new search API allows AI tools to retrieve Slack data without storing it externally.

Slack also introduced tighter permission scopes to give enterprises more precise control over what AI can access in private channels and direct messages.

Anthropic Ships Claude Sonnet 4.6

Anthropic released Claude Sonnet 4.6, which delivers Opus-level performance at Sonnet pricing ($3 input / $15 output per million tokens). The model scores 72.5% on OSWorld-Verified, nearly double its predecessor, and supports a one-million token context window in beta. Developers prefer it over both Sonnet 4.5 and Opus 4.5 in head-to-head comparisons.

It is now the default model in Claude Code and includes new MCP connectors for the Claude in Excel add-in, allowing agents to pull data from sources like FactSet and Moody's directly into spreadsheets.

‍

Vendor / Product	Key Action / Feature	Significance
Anthropic and Infosys	Strategic partnership combining Claude models with Infosys Topaz to build enterprise agents for telecom, financial services, and manufacturing	Targets regulated industries where governance and auditability are non-negotiable; includes a dedicated telecom Center of Excellence
Anthropic and Rwanda	Three-year government MOU covering health, education, and public sector AI across eight African nations	Marks Anthropic's first multi-sector government partnership on the African continent; provides Claude Code access to public sector developers
Anthropic and CodePath	Integration of Claude and Claude Code into core computer science curricula for 20,000+ students at community colleges and HBCUs	Targets economic mobility and builds a pipeline of agent-native engineers from underrepresented communities

Google Updates

This week, Google significantly improved its model and infrastructure capabilities with releases focused on frontier reasoning, autonomous coding, and improved enterprise data access.

Product	Key Update	Significance
Gemini 3.1 Pro	77.1% on ARC-AGI-2, 80.6% on SWE-Bench Verified; 1M-token context; preview across Vertex AI and Gemini Enterprise	Positions Google as competitive in autonomous coding and long-context enterprise workflows
Gemini 3 Deep Think	84.6% on ARC-AGI-2; 3455 Codeforces Elo; early access API	Signals Google’s push into high-end scientific and engineering reasoning
Google Cloud MCP Expansion	Managed MCP support across AlloyDB, Spanner, Cloud SQL, Firestore, and Bigtable	Gives agents unified, secure access to operational databases across Google Cloud

OpenAI Hires Creator of OpenClaw to Lead Personal Agents

OpenAI hired Peter Steinberger, the developer behind OpenClaw, to lead its personal agents division. OpenClaw reached nearly 200,000 GitHub stars and 1.5 million agents since launching in November 2025. Steinberger will build autonomous systems that manage emails, calendars, and digital workflows without manual input, while OpenClaw transitions to an independent open-source foundation with OpenAI backing. CEO Sam Altman described the company's direction as "extremely multi-agent."

Microsoft Boosts AI Security with New Security Dashboard

Microsoft has significantly enhanced its AI security offerings by launching the Security Dashboard for AI in public preview and publishing its internal MCP security blueprint. The new Security Dashboard provides security teams with a unified view of risks across their entire AI estate, including Microsoft-native tools and third-party platforms like OpenAI and Google Gemini, by consolidating signals from Defender, Entra, and Purview to surface threats like data leaks and model drift.

Other Updates

Vendor / Product	Key Action / Feature	Significance
Manufact (formerly mcp-use)	Raised $6.3M seed round led by Peak XV with participation from Y Combinator to scale its MCP infrastructure platform	Addresses the connectivity and deployment gap for developers building MCP-compliant agents; reports over 6 million SDK downloads
Autosana	Secured $3.2M seed funding to scale an agentic QA platform for mobile and web testing	Replaces brittle test scripts with AI agents that self-heal when UI elements shift, reducing manual regression testing for fast-moving teams
Redpanda	Launched new core capabilities for its Agentic Data Plane, including a centralized AI Gateway for governance and MCP server access control	Provides enterprises with unified authentication, real-time observability, and a policy layer across all agent interactions with live data

My Thoughts: Why Enterprise AI Is Shifting to Security-First Infrastructure

The focus has shifted from raw AI model capability to the reliable and secure deployment of AI agents. New infrastructure (Slack, Microsoft) and platforms (Redpanda) emphasize security, auditability, risk control, and governance to safely integrate AI into business workflows. The primary challenges are now trust, security, and scale, not intelligence alone.

AI deployment and security will see three major shifts: Security and governance will become the primary investment focus due to the need for auditability; "reasoning wars" will intensify, making complex problem-solving the enterprise standard as simple generation is commoditized; and AI will become a global infrastructure layer, with models like Claude Sonnet 4.6 accelerating sovereign and public sector adoption.

‍

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