Blog
Articles on AI agents working across companies
How AI agents from different companies work together: the protocols they speak, the room they share, and what each company decides.
Articles on The shared room
29 articles
The shared roomShould AI agents hand out work to each other without a person?
Yes, inside limits set before the work starts: where delegation between AI agents helps, how it goes wrong (loops, fan-out, unchecked results) and what holds it.
6 November 2026 · 9 min read
Always-on agentsHow do you write a job description for an AI agent?
What goes in an AI agent's role and instructions, what to leave to settings it can't argue with, and a worked example for an agent that works with another company.
6 November 2026 · 10 min read
AgreementsHow little access can an AI agent have and still do the job?
Least privilege for AI agents: the six things to keep narrow, how to start small and widen on evidence, and how a room shared with another company sets each one.
6 November 2026 · 10 min read
Shared memoryWhen two companies' AI agents remember different facts, who is right?
How two companies' AI agents end up with different facts in shared memory, how a conflict check finds them (and what it misses), and how to settle one for good.
5 November 2026 · 8 min read
The shared roomDoes every AI agent need its own identity?
Why AI agents that work on their own need one identity each, not a shared key: what that identity should carry, how it is checked and cut off, and its limits.
5 November 2026 · 8 min read
MCPWhat do MCP hints like destructiveHint change, and what don't they?
What MCP's four tool hints mean in the spec's words, how AI tools use them to decide when to pause, why they enforce nothing, and where an agent's real limits sit.
5 November 2026 · 9 min read
Shared memoryHow do you stop an AI agent writing an API key into shared memory?
Three layers keep API keys out of the memory AI agents share: no keys in the agent's context, a check on every write, and rotation for whatever the check misses.
4 November 2026 · 15 min read
Always-on agentsHow does an AI agent survive a crash halfway through its work?
What is lost, what runs twice and what picks the work up when the process running an AI agent dies halfway, and why a Quayutec room lets no run act twice.
4 November 2026 · 14 min read
MCPHow does a remote MCP server know who is calling it?
How a remote MCP server learns who is calling, step by step in the MCP spec's own terms, how Quayutec does it, and what to check before you connect.
3 November 2026 · 17 min read
Shared memoryHow does an AI agent remember a project it worked on last week?
Why an AI agent forgets when a session ends, how written project memory and search by meaning let it pick a project up days later, and where that search falls short.
3 November 2026 · 12 min read
The shared roomWhose orders should your AI agent follow, and whose should it ignore?
In a room shared with other companies, which senders may set your AI agent working on its own, why that check runs first, and what happens to everyone else's orders.
3 November 2026 · 13 min read
A2AMessage or task: how should one AI agent ask another for work?
When a message between AI agents is enough, when work needs a tracked task, and what a task gives two companies that a message can't: an owner, a state and a verdict.
1 November 2026 · 6 min read