Tentra is an AI-native architecture platform for engineering teams. Describe a payment system with Stripe, Kafka, and PostgreSQL in natural language — get an interactive diagram, then export to Java Spring Boot, Python FastAPI, Go, Rust, or 10 other frameworks in 30 seconds. Drift detection catches when your diagram falls out of sync with real code. Built inside Cursor, Claude Code, Codex, and Windsurf via MCP.
You describe: "Design a payment system with Stripe, Kafka, and PostgreSQL." → Interactive diagram: 5 services, 6 connections, typed edges. → Java Spring Boot scaffold (or 13 other frameworks). → Drift detection runs on every commit.
AI generation from natural language plus 14-framework code export plus drift detection. No DSL to learn, no diagrams to hand-draw. Zero LLM cost on our side — the agent in your IDE does the semantic work.
Drawing architecture in Lucidchart or draw.io eats hours — drag, connect, align, repeat for every new service.
Structurizr makes you learn a DSL before you can contribute a single diagram.
IcePanel at $40–80 per editor per month still can’t generate a diagram from a sentence or export a scaffold.
Diagrams rot the moment code changes — by the time the PR merges, docs are already lying.
AI generation from natural language: describe your system in a sentence, get an interactive diagram with typed services and connections.
14-framework code export: Java Spring Boot, Python FastAPI, Go chi, Rust Axum, .NET, Kotlin Ktor, Ruby Rails, and more — real project scaffolds, not boilerplate generators.
Drift detection: sync_architecture compares saved diagrams against your live codebase and reports exactly what changed.
Agent-as-LLM: your IDE’s own model (Cursor, Claude Code, Codex, Windsurf) does the semantic work. Zero LLM cost on our infra. Zero API-key setup on yours.
Tentra also ships a persistent code-graph layer so AI coding agents can query symbols, call graphs, and architectural hotspots across sessions. Free in local mode via npx tentra-mcp --local init. The graph stores files, symbols, imports, call edges, ADRs, contracts, and ownership. Secondary to the architecture workspace above.
Sign in with GitHub, get your API key, add the SSE endpoint to your IDE. Zero install — just a URL. Works inside Cursor, Claude Code, Codex, Windsurf via MCP.
What is Tentra? Tentra is an AI-native architecture platform for engineering teams. Describe a system, get an interactive diagram, export to 14 frameworks, detect drift.
How does AI architecture generation work? Describe "Design a payment system with Stripe, Kafka, and PostgreSQL" in your IDE. Your agent calls create_architecture, returning a typed service topology with connections on a shareable interactive canvas.
What frameworks can Tentra export to? 14 frameworks: Java Spring Boot, Node.js Fastify, Python FastAPI, Go chi, Rust Axum, .NET ASP.NET Core, Kotlin Ktor, PHP Laravel, Ruby Rails, Elixir Phoenix, Docker Compose, Terraform, Mermaid, ADR.
What is drift detection? sync_architecture compares your saved diagram against the live codebase and returns an accuracy score 0-100 plus a structured diff. Run it in CI or during PR review.
How does Tentra compare to IcePanel, Lucidchart, Structurizr, Eraser? Tentra does everything they do (diagrams, cloud components, C4-style topologies) plus AI generation from natural language plus 14-framework code export plus automatic drift detection.
Do I need an OpenAI or Anthropic API key? No. Tentra uses the agent-as-LLM pattern — your IDE's agent does the semantic work. Zero LLM cost on our side.
Is Tentra free? Yes — Tentra is currently in free beta. All features are available at no cost. No credit card required.