Loading…
FIELD NOTES
Notes from the agent control plane.
Working write-ups from the engineers at Bytevon — how we govern agent-to-agent trust, turn goals into observable plans, secure the systems we build, and keep all of it in plain sight. No thought-leadership. Just how the systems we ship actually run.
ENGINEERINGA2A POLICYPLAN AUTOMATIONAGENT SECURITYOBSERVABILITY
FEATURED
A2A POLICY
Policy gates, not blind handshakes: governing agent-to-agent trust
When agents delegate to other agents, an implicit handshake is a liability. We walk through machine-readable A2A policy — discovery, trust negotiation, and enforceable delegation at the agent boundary.
ALL POSTS
ENGINEERING
RAG or fine-tuning? A decision guide for agent knowledge
Two ways to give an agent domain knowledge, constantly confused for each other. Here is the practical decision tree we use on real projects — and why retrieval wins more often than teams expect.
7 min read
A2A POLICY
MCP in practice: exposing internal tools to agents without losing control
The Model Context Protocol lets any agent use your tools through one interface. That convenience is also a risk surface. Here is how we expose MCP servers behind policy, scope, and audit.
8 min read
PLAN AUTOMATION
Plans over autonomy: making agent work observable and revisable
End-to-end autonomy is a black box no one can audit. We make the case for structured plans — observable, pausable, and revisable mid-execution — as the unit of agent work.
7 min read
AGENT SECURITY
On-prem agentic AI: a reference architecture for regulated industries
When data cannot leave the building, cloud AI is off the table — but agents are not. Here is the fully on-prem, open-source architecture we run for a regulated engineering customer.
9 min read
AGENT SECURITY
Every agent an identity: the security model for a 1,000-agent fleet
Fleets introduce a new attack surface: compromised agents, privilege escalation, and opaque tool use. Here is the identity, sandboxing, and immutable-audit model we run.
8 min read
OBSERVABILITY
Telemetry from every gate: observability for agent fleets
You cannot operate what you cannot see. We cover the structured telemetry we instrument into every agent, plan step, and policy gate — and how a team steers a live fleet from it.
6 min read