Driving pi with computer using ChatGPT sessions
How I use ChatGPT tasks to organize persistent Pi sessions, connect MCP tools, bridge gaps with computer use, and deliver work without losing context.
16 stories
How I use ChatGPT tasks to organize persistent Pi sessions, connect MCP tools, bridge gaps with computer use, and deliver work without losing context.
AI coding agents put internal screenshots from more than 300 organizations on public GitHub. No model broke. The agents used permissions nobody meant to give them, to finish a task nobody thought was risky.
igma's remote MCP server admits clients by the client_name they send during dynamic client registration. Why that check proves nothing, what CIMD fixes and what it doesn't, and what to gate on instead.
The MCP Python SDK, the Rust SDK (rmcp), and LiteLLM each accepted authentication input that nobody verified. Here is what broke, who is affected, and the checklist that closes the gap.
Diagnose failed assistant actions across host, MCP tool, backend, and source of record with privacy-conscious evidence, safe replay, and a ticket template.
ChatGPT is adding draft MCP Events support. Why a subscription is a credential, how revocation works, and what your MCP server must store.
Compare the App Store’s launch with OpenAI plugins across discovery, trust, payments, portability, and gatekeeping, and see where the analogy fails today.
Cloudflare, Okta, Auth0 and Microsoft all offer MCP gateways now. A side-by-side comparison of where each one sits, how it handles auth and logging, and what your MCP server still needs to do.
How workos mcp install configures Claude Code, Codex, and Cursor, why configured is not the same as authenticated, and the failure modes worth knowing.
A practical, testable list of controls for letting AI agents act on your users' data without handing them the keys.
A first look at how the official Model Context Protocol SDK is turning the 2026-07-28 spec's authorization hardening into code, and what it means for anyone running an MCP server.
CoPhish showed that an agent-building platform can itself become an OAuth phishing vector. Here's what that means for anyone building MCP servers or agent platforms, how to configure against it, and what Entra Agent ID and WorkOS Agent Auth reveal about where the industry is headed.
Two critical vulnerabilities, one root cause, and the piece of MCP that is supposed to close this gap.
A new registry audit counted servers, not security controls. The number matters more than it sounds like it should.
The same token type is solving two unrelated problems. Knowing which one you are in decides what you build.
The case against MCP is right about ergonomics and quiet about authorization. Here is what has to exist before direct API access is a safe default.
Showing 16 of 16 stories
You're all caught up.
Get the top AI news and tools straight to your inbox every week.
Weekly · AI news from WorkOS Blog
Join 15,000+ AI enthusiasts