MongoDB Atlas

Every collection your services actually read and write.

MongoDB joins the graph through OpenTelemetry — instrument your app once and Neat mints a live edge for every collection your services actually touch. No connector add; the graph fills from the mongoose spans your app emits.

Paste it into your coding agent — it runs the full setup with your credentials.

Wire my MongoDB (Atlas or self-hosted) into NEAT. MongoDB is an OBSERVED connector — it enters the graph through OpenTelemetry, not `connector add`. NEAT mints a live edge for every collection my services actually read or write, from the mongoose spans the app emits. Steps:
1. My app must use MONGOOSE. The raw mongodb driver's OTel instrumentation does NOT emit command spans on modern drivers (6.x/7.x) — only mongoose instrumentation produces the db.mongodb.collection span NEAT needs. If I'm on the raw driver, tell me — that's a blocker.
2. Add OpenTelemetry auto-instrumentation: install @opentelemetry/sdk-node @opentelemetry/auto-instrumentations-node @opentelemetry/exporter-trace-otlp-http, and load an init module (--require for CJS, --import for ESM) that starts NodeSDK with getNodeAutoInstrumentations() and an OTLP exporter pointed at my daemon's http://<daemon>/v1/traces. Use a SimpleSpanProcessor for short-lived scripts so spans flush before exit.
3. Run the app so it performs real DB operations.
4. Verify: check the graph for infra:mongodb-collection:<name> nodes with OBSERVED CALLS edges from my service (mongoose pluralizes model names, e.g. Order -> orders). Report the collections observed, and flag any two services writing to the same collection with different field names.

how it works

Spot two services writing different field names to the same collection — a bug no single repo would show.

Add OpenTelemetry auto-instrumentation to your app — mongoose spans are what Neat reads.
Run the app so it performs real DB operations. The spans flow to your daemon.
Each collection appears as infra:mongodb-collection:<name> with OBSERVED edges from every service that touches it.

Other connectors

Supabase The tables and RPCs your app actually queries in production. Vercel Your production traffic's traces, pushed into the graph. Railway Services and deploys, no app code required. Cloudflare Which Workers actually run in production, mapped back to your code.