What Is an Agent Harness? The Part Everyone Skips
The five parts of an agent harness: capability surface, meaning, policy, approval and audit. Only two of them are about power.
Agent Harness01
Series · 4 parts · 2,025–2,026
Agent Harness
The layer between a language model and software that already exists: what to expose, how to keep it safe, and how to store an agent so it can be changed without a deploy.
Give the Model Your Verbs, Not Your Tables
Why intent-named operations beat database schemas, auto-generated OpenAPI specs and one-command MCP servers when you hand an application to a model.
Agent Harness02
Safe by Default: Building Agents That Can't Wreck Your Data
Read/write enforcement, principal scoping, approval gates, sandboxes and budgets: five safety properties, every one enforced in code, not in a prompt.
Agent Harness03
An Agent Is Data, Not Code
Store agents as versioned, validated rows instead of writing them as classes, and rollback, diffs, audit, teams and lessons come free. Here is the bill.
Agent Harness04