Definitions for the primitives behind agents, workflows, memory, tools, and safety.
Documentation
What is BuildAI?: Understand BuildAI as an agent infrastructure platform, not a chatbot builder.
Governed execution: How the action gate decides every consequential step: risk classes, exact single-use approvals, separated duties and the model proposer.
Release governance: Versioned release policies, separate duties for policy, promotion and override, and customer outcome thresholds.
Graders and evaluator authority: Deterministic checks, human review and calibrated model judges, plus who may grade, and how evaluation data is retained and purged.
Model registry: Approve or block provider models with evidence, and opt into strict mode so only approved models run.
Operations and incidents: Operational measures with denominators and intervals, an append-only incident timeline, and member suspension.
AI Agent Infrastructure: The core layers behind reliable agents, workflows, tools, memory, and safety.
The BuildAI AI Harness: Define the operating layer that connects models to context, tools, policy, evaluation, and evidence.
Inference Harness and Model Gateway: Understand provider configuration, capability discovery, model access, latency, cost, and workspace policy.
The BuildAI Evaluation Harness: Evaluate model and agent capability, safety, human preference, operations, and task completion with evidence.
Observability and Economics: Connect traces, quality signals, latency, usage, cost, and outcome evidence without inventing unavailable data.
BuildAI System Design: A clear end-to-end architecture for surfaces, gateway, orchestration, runtime, and logs.
Enterprise Control Plane: How BuildAI connects identity, policy, providers, approvals, usage, and audit inside one workspace boundary.
Security and Governance: Tenant isolation, credential boundaries, human review, auditability, and production-safe operating principles.
Core Concepts: Definitions for the primitives behind agents, workflows, memory, tools, and safety.
BuildAI for Different Users: How individuals, founders, businesses, developers, agencies, and consultants use BuildAI.
Example Agent Workflows: Concrete workflow patterns you can adapt into agents and task templates.
Developer APIs: Repository-backed API reference, authentication boundaries, envelopes, and errors.
Capability availability: The current status of every BuildAI capability, sourced from the product repositories.
Glossary: Definitions for the agent infrastructure terms used throughout BuildAI.