Skip to news
English

AI & MCP News — 2026-10-26

7 verified updates published on 2026-10-26. Each summary links to an official announcement or release.

← All news

Vercel Sandbox adds stdin support for MCP servers

Vercel Sandbox can now send standard input to running commands, enabling exchanges with stdio MCP servers. Its TypeScript and Python SDKs support writing input and closing the input stream.

Published · Source checked

Vercel · Official source ↗ # Permalink

Vercel Connect adds service discovery from the CLI

Vercel CLI 64.1.0 and later lets users and agents search more than 1,000 Connect services from the terminal. Results support MCP or OAuth connection filters and JSON output for automation.

Published · Source checked

Vercel · Official source ↗ # Permalink

Oracle introduces an OCI Observability MCP server

Oracle introduced an MCP server for querying observability data, correlating events and generating charts and incident reports. Access follows its configured OCI identity, with an optional read-only mode and separate confirmation for guarded resource changes.

Published · Source checked

Oracle · Official source ↗ # Permalink

OpenAI releases GPT-6.1 Terra in its API, Work and Codex

GPT-6.1 Terra is available through the OpenAI API and to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex. Standard API prices per million tokens are $1 for input, $0.05 for cached input and $5 for output.

Published · Source checked

OpenAI · Official source ↗ # Permalink

Halon adds read-only MCP access to email operations

Halon Engage and Halon Protect 26.3 introduce Halon's first MCP integration, letting agents investigate email delivery and security data. Initial access is read-only, preventing configuration changes and actions on queued messages.

Published · Source checked

Halon · Official source ↗ # Permalink

Google introduces IAM Helper skills and managed MCP tools

Google introduced IAM Helper Agent Skills for permission troubleshooting, policy management, policy simulation and temporary privileged access. They use read-only tools exposed through Google-managed MCP servers; access-policy changes require human confirmation.

Published · Source checked

Google Cloud · Official source ↗ # Permalink

Devin Security adds enterprise deep scans using Claude Mythos

Cognition upgraded Devin Security with enterprise deep scans powered by Claude Mythos and other cyber models, without requiring separate model access. It offers free scans to current and future enterprise customers; new customers can also receive remediation credits.

Published · Source checked

Cognition · Official source ↗ # Permalink