Five Phases, Human Control at Every Gate

Each phase pairs agent capability with human judgment. Agents produce output; humans evaluate coherence. The pipeline adapts to your team's trust level — early on, humans review everything. As trust builds, the review surface shrinks.

Discovery

Agents analyze requirements, surface assumptions, and map the problem space. Humans set boundaries and priorities.

Specification

Agents draft specs from discovery output. Humans review for coherence with existing architecture and business intent.

Workflow

Agents generate implementation plans. Humans approve scope and verify the plan is achievable given team constraints.

Execution

Agents produce artifacts — code, tests, documentation. Humans audit at policy boundaries defined per work cycle.

Audit

Agents run coherence checks across the graph. Humans review flagged incoherences and decide whether to accept, revise, or roll back.

The Product Model

Computed Coherence, Not Status Meetings

The graph knows which artifacts depend on each other. When something changes, the engine computes what's affected. Your team reviews the report instead of playing detective across Jira, Confluence, GitHub, and Slack.

Staleness Propagation

When an upstream artifact changes, the propagation engine walks the dependency graph and marks downstream artifacts as stale — surfacing what needs review without declaring it broken.

Version Tracking

Every node in the graph carries version fields. The product model compares versions across edges to detect drift between related artifacts.

Coherence Reports

A computed summary of what's in sync, what's stale, and what the team should prioritize. Replaces status meetings with data.

Graph Schema

FeatureNodes, SpecNodes, WorkflowNodes, and WorkCycles connected by typed relationships. Queryable via Kuzu graph database.

Environment Templates

Eight Starting Points, Infinite Customization

No two teams are alike. Codesign ships with 8 environment templates that configure pipeline gates, review cadences, and coherence thresholds for different organizational shapes.

Startup (lean team, rapid iteration)

Enterprise (compliance gates, approval chains)

Agency (multi-client, parallel streams)

Open Source (community contributors, async)

Research (experiment-driven, hypothesis tracking)

Platform (API-first, multi-tenant)

Migration (legacy modernization, incremental cutover)

Hybrid (mixed methodology, transition state)

Under the Hood

Technical Architecture

Runtime

Python with FastAPI

Graph Database

Kuzu (embedded, zero-ops)

Deployment

Docker Compose

LLM Routing

Multi-provider (Anthropic, OpenAI, Gemini, Grok, Ollama)

Connector Pattern

Any language, any framework, any methodology

Data Sovereignty

Your API keys, your infrastructure, your data