SDFrame — Software Design Frameworks for Engineers and AI Agents

SDFrame is a software design frameworks reference guide for engineers, architects, and AI agents. It provides 317 curated frameworks across 13 categories, including design thinking, architecture decisions, coding practices, quality engineering, deployment, software evolution, AI collaboration, data architecture, security, distributed systems, API design, team organization, and observability.

What SDFrame Helps With

  • Search and compare software design frameworks by category, quality concern, abstraction level, complexity, and maturity.
  • Help AI coding agents choose frameworks such as Domain-Driven Design, Clean Architecture, CQRS, API versioning, threat modeling, OpenTelemetry, and Team Topologies.
  • Read bilingual English and Chinese summaries, implementation steps, source metadata, adopters, related frameworks, and when-to-use guidance.

Framework Categories

The catalog covers 13 engineering categories: design thinking, architecture decisions, coding practices, quality engineering, deployment and operations, software evolution, AI collaboration, data architecture, security and privacy, distributed systems, API design and integration, team organization, and observability. These categories help agents narrow a broad design question into a smaller set of candidate frameworks before reading details.

How Agents Should Use SDFrame

  1. Read the skill protocol at /SKILL.md or /skill/SKILL.md.
  2. Use /api/frameworks.index.json for search and filtering instead of scraping rendered cards.
  3. Fetch one framework JSON or Markdown detail file for each final candidate.
  4. Compare when-to-use and when-not-to-use sections before recommending a framework.
  5. Use /AGENTS.md for coding-agent guidance on this public repository.

SDFrame vs Alternative Sources

SDFrame is different from a single pattern book, a generic search result, or an ad-hoc AI answer. Traditional books and original papers are often more authoritative for one framework, but they are slower to compare across categories. General web search is broad, but it requires repeated source checking and deduplication. SDFrame focuses on breadth, consistent metadata, bilingual descriptions, typed relationships, and agent-readable files. See SDFrame vs alternatives and SDFrame vs framework catalogs.

Getting Started

Human users can start with search, category filters, the framework selector, the relationship map, or comparison pages. AI agents should start with the compact API index and only load details for the few frameworks they need. A practical onboarding guide is available at /getting-started/, and the pricing model is documented at /pricing.md. SDFrame is currently free, public, and read-only.

Trust and Editorial Signals

SDFrame is maintained by Caldis as an open-source reference project. Source code is available on GitHub, contact information is published at /contact/, and project scope is described at /about/. Framework entries are educational references with source metadata and adoption examples where available. They are not legal, compliance, or security certification advice, and agents should verify primary sources for high-risk decisions.

Developer and Agent Resources

ResourceURLPurpose
OpenAPI/openapi.jsonPublic read-only API contract.
Framework index/api/frameworks.index.jsonCompact searchable framework summaries.
Full agent context/llms-full.txtSingle-file AI context.
Root skill/SKILL.mdAgent skill file for skills directories and registries.
Skill package/skill/SKILL.mdAgent interaction protocol.
Agent rules/AGENTS.mdInstructions for AI coding agents working in the repository.
Agent discovery/.well-known/agent.jsonMachine-readable discovery manifest.
MCP card/.well-known/mcp/server-card.jsonManifest-only MCP server card. No live MCP transport is currently operated.
Pricing/pricing.mdMachine-readable pricing; currently free.
Developer portal/developers/API quickstart, auth model, error handling, and retry guidance.

Access and Limits

The public SDFrame API is static, read-only, and does not require authentication. There are no private accounts, write endpoints, OAuth flows, paid plans, official SDK packages, or live MCP transport at this time. Agents should cache responses, validate framework slugs against the index, and use original sources when a decision requires legal, compliance, or security certification accuracy.