How Archynt works
Four surfaces, one graph — from static analysis to live topology, custom metrics dashboards and AI-powered insights.
Projects
Understand any codebase as a living architecture. Archynt parses Java, Node, Python and Go and builds the graph automatically.
Runtime
A read-only agent discovers live services, dependencies and host pressure — zero instrumentation required. Docker today; Kubernetes and systemd are next.
Metrics
Build Grafana-style dashboards from your Prometheus metrics with free-form PromQL, thresholds and org-scoped panels.
AI insights
An LLM reasons over your architecture graph — static or live — to surface coupling issues and god-services, with evidence and a concrete fix.
Frequently asked
How is Archynt different from Datadog, Grafana or Dynatrace?
Those platforms answer "what is happening right now" from telemetry — metrics, logs and traces. Archynt answers a different question: "is the system that's actually running the one the code describes?" It parses source code into a structural graph, correlates it with a read-only runtime agent, and tags every dependency as static, runtime, or both. Archynt can ingest a team's existing Prometheus metrics to enrich that graph, but it doesn't compete on telemetry storage — its unit of value is the architecture graph and the reasoning over it, not dashboards.
Does Archynt see my source code?
Static analysis parses the repository to extract classes, modules, endpoints, and their relationships into a graph — it does not ship your source text to a third party for storage. The runtime agent is read-only and never modifies or executes anything on the observed system.
What is architectural drift?
Drift is a dependency that exists in one layer but not the other: a service call observed in live traffic that no code path documents, or a dependency declared in code that the runtime never actually exercises (a dead-code candidate). Archynt's confirmed_by tag on every edge — static, runtime, or both — is what makes drift detectable at all.
What languages and frameworks does Archynt analyze?
Java/Spring, Node/NestJS, Python and Go today, via dedicated per-language workers behind one graph model. More languages are added by extending the same worker contract.
Do I need to instrument my application?
No. The runtime agent discovers services, containers and live dependencies from the Docker API and /proc — no SDKs, sidecars, or code changes. RED metrics (rate, errors, duration) come from an eBPF sensor (Beyla), also without touching application code.
Can AI coding agents use Archynt directly?
Yes. Archynt exposes an MCP server so tools like Cursor or Claude Code can query the real architecture graph, list open design problems, and check whether a proposed change crosses an architectural boundary before generating code. See /llms.txt and /.well-known/mcp.json for the machine-readable description.
Ready to try it?
Run your first analysis or jump into your organization workspace.