Writing
#agents
Every post tagged agents.
Accountability has to spread
The risk from autonomous agents does not sit inside the frontier labs. Agents and their subagents already run on many machines across many organizations, so accountability cannot be something the frontier labs bolt on for us. It has to be local-first and portable, installed where the agent runs, and checkable by strangers.
What Zerker Gateway Can Be
We've gotten very good at building agents and very bad at building the plumbing between them. Zerker Gateway is the traffic path for AI agents -- self-hosted, in the path, and honest about every call.
Capability Cards: Proving What an Agent Can Do, Not Just What It Says
Descriptor formats like A2A's AgentCard tell you what an agent claims it can do. None of them tell you whether the claim is bound to a key you trust, or whether the agent's actual behavior matches. Here is the arc we shipped to close that gap: a predicate registry, signed capability cards, per-actor signing that makes an agent's identity provable, revocation, and the same verdict in the browser as on the command line.
A2A Makes Agents Interoperable. Treeship Makes That Interoperability Trustworthy.
Google's Agent2Agent protocol gives every agent a way to talk to every other agent. It does not give you a way to verify what any of them actually did. Here is how @treeship/a2a closes that gap, with worked examples you can run end-to-end.
Every MCP Tool Call Your Agent Makes Right Now Has No Receipt
The Model Context Protocol specification is explicit: tool calls are arbitrary code execution and hosts are responsible for authorization. Most MCP implementations have no answer to this. Here's what closing that gap looks like.
The Key Management Nightmare Hiding Inside Agent Attestation
Teams adopting cryptographic attestation for agents keep hitting the same wall: one keypair per agent sounds right until you have forty agents and no coherent governance. Here's the model that actually works.
The Replay Attack Your Authorization System Doesn't Prevent
Most authorization systems for AI agents share a subtle flaw: approvals can be reused. An agent that captures an approval token can replay it. Here's the attack and how approval-based authorization prevents it by construction.
Chain of Custody for AI Agents: What Software Can Learn from Physical Evidence Handling
Physical evidence handling has solved chain of custody over centuries. AI agent workflows need exactly this.
Why Agent Actions Need Receipts
When a human takes an action, there's context: intent, memory, accountability. When an agent takes an action, there's just a log line. That asymmetry is the problem Treeship solves.