X MCP Adds Read-Only Mode for Safer Agents

// 21m agoPRODUCT UPDATE

X MCP Adds Read-Only Mode for Safer Agents

X MCP now supports ?read_only=true, removing posting, DMs, bookmarks, blocks, and other write tools while rejecting write attempts even when tokens retain write scopes. The mode works across X API and Ads API MCP servers.

// ANALYSIS

This is a practical safety boundary for agents connected to high-impact social accounts, though it complements rather than replaces least-privilege credentials and human approval.

–Read-only mode makes research and monitoring safer by removing accidental side effects
–URL-level capability control is easy to deploy across MCP clients
–Server-side rejection matters because prompt instructions alone cannot reliably prevent writes
–Developers still need a separate, carefully scoped configuration for legitimate posting or messaging workflows

// TAGS

x-mcpmcpagenttool-useapisecurityautomation

DISCOVERED

21m ago

2026-11-10

PUBLISHED

1h ago

2026-11-10

RELEVANCE

8/ 10

AUTHOR

XDevelopers

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

UPDATE1h ago

Cursor Projects adds multiplayer agent teams

Cursor has rebuilt Projects so teammates can share a task board and persistent context while their coordinator agents message one another directly. The feature requires a team account and Cursor 3.26.13 or later.

UPDATE1h ago

Orca Adds Custom App Icons in Settings

Orca, a desktop IDE for running multiple AI coding agents in parallel, now lets users import a custom app icon by searching “icon” in Settings. The feature arrived in the 1.4.254 update alongside broader interface polish.

OPEN SOURCE1h ago

Offense Brings Security Tools to Coding Agents

Offense is an open-source security toolset that gives coding agents CLI, MCP, and HTTP access to dependency audits, source analysis, workflow checks, and remediation verification. Offense Research offers optional hosted execution alongside local and self-hosted modes.