Search topic: AI agent harness
AI agent harness architecture
Definition, components, execution protocol, controls, and a production implementation path.
Read the primary guideField guides
Engineer reliable agent work
Start with one workflow and an observable definition of done. These guides explain context, tools, authority, evaluation, and recovery. Use the deeper references when a measured failure calls for them.
Start with the harness adoption guideInteractive visual guide
One request. Every boundary visible.
Follow a refund through evidence, permission and completion. Change the conditions to see where a reliable harness holds, denies or reconciles the run.
Search topic: AI agent production readiness checklist
AI agent production readiness
A practical checklist for context, tools, identity, policy, evaluation, replay, and operations.
Read the primary guideSearch topic: AI agent harness audit
AI agent harness audit
A repository-based assessment of the production controls around a tool-using AI agent.
Read the primary guideSearch topic: AI agent context pack
Context Pack for AI agents
The versioned schema, compiler contract, worked example, and lifecycle for bounded runtime context.
Read the primary guideSearch topic: AI agent evaluation framework
AI agent evaluation and scorecards
Datasets, policy and utility checks, release gates, production slices, and regression replay.
Read the primary guideSearch topic: AI agent memory architecture
AI agent memory architecture
Working, episodic, semantic, procedural, and organizational memory with governed promotion.
Read the primary guideSearch topic: AI agent security best practices
AI agent security threat model and controls
Source-to-sink threat modeling, scoped identity, tool containment, approval boundaries, security evals, and audit evidence.
Read the primary guideSearch topic: MCP security
MCP security for production agents
Capability manifests, schemas, authorization, approvals, idempotency, observability, and replay.
Read the primary guideSearch topic: context engineering for AI agents
Context engineering for production AI agents
Versioned context assembly, evidence selection, token budgets, policy, and replayable compilation.
Read the primary guideSearch topic: multi-agent orchestration architecture
Multi-agent orchestration architecture
Specialist lanes, bounded delegation, durable sessions, authority, and final decision ownership.
Read the primary guide