10 design documents.
One source of truth.

The PRD. Problem, evidence, goals, non-goals, competitive posture, pricing surface, year-1 plan.

For: Customers, design partners, investors

How the OSS library is built. Monorepo layout, package boundaries, harness primitives, framework-agnostic core.

For: OSS contributors, library consumers

The SaaS architecture on AWS. Control plane, data plane, multi-tenant isolation, GitHub App, billing.

For: Platform engineers, infra reviewers

The 3-pass LLM pipeline. Model routing, prompt caching, harness-primitive grounding, confidence scoring, eval harness.

For: ML infra engineers, applied AI leads

STRIDE threat model, SOC 2 Type II control mapping, customer-cloud runner, encryption, IR plan.

For: Customer security teams, auditors

SLOs, on-call rotation, deployment pipeline, capacity planning, runbooks, backup/DR, GA milestones.

For: SRE, on-call engineers

The moat doc. Per-tenant data assets, RAG retrieval, weekly tuning cycle, cross-tenant insights. Why customers pay.

For: Platform engineers, prospective customers

Personas, end-to-end journey, dashboard wireframes, CLI, GitHub App, notifications, onboarding, RBAC, accessibility.

For: PM, design, customer success

Per-vendor integration spec for Jam, LogRocket, Sentry Replay, FullStory, Datadog RUM. Endpoints, payloads, normalization, quirks.

For: Adapter implementers, vendor partners

The closed-loop input edge. Why first-party beats third-party for the agent loop, MV3 architecture, captured event surface, programmatic API, three integration patterns including local-agent endpoint and MCP server roadmap.

For: Engineers integrating the recorder, agentic-coding integrators

What the SaaS centralizes — the org-wide queryable corpus of sessions, specs, runs, bug-classes, and signals. Six entity classes, REST API, eight-tool MCP server for Claude Code / Cursor / Aider mid-task investigation. The OSS-vs-SaaS buy-vs-build line.

For: Engineering leaders evaluating the SaaS, integration engineers