RetrySight You shouldn’t need a sales call to figure out which version of a dev tool to install. RetrySight ships as three deployment models that share the same retry taxonomy and collector architecture — the difference is who’s looking at the data and where it’s allowed to live, not what it measures.
Start with one question
Who needs to see this data?
- Just you, on your own machine → Lite
- Your team, and you’re fine with a managed console → Cloud
- Your org, and telemetry can’t leave your network → Enterprise
You can start on Lite and move to Cloud or Enterprise later without losing the mental model — collectors still read the same local agent logs either way.
RetrySight Lite — solo, local-only
Free and open source (Apache 2.0). A native desktop app for macOS, Windows, and Linux with in-process collectors — no Docker, no fleet manager, nothing to deploy. REST, GraphQL, and MCP are all served on loopback (127.0.0.1:18081), and nothing uploads anywhere unless you point it at a remote manager yourself.
Right fit if you’re an individual developer who wants to see your own retry patterns without asking a platform team to stand up infrastructure first. Download RetrySight Lite.
RetrySight Cloud — team, managed
We host the manager and dashboard; your developers’ laptops run a lightweight collector that pushes to it. Per-account ingest keys, row-level multi-tenant isolation, and the same retry taxonomy and cost views as Enterprise — without standing up your own stack. Currently in early access at cloud-staging.retrysight.com, so expect the environment to keep evolving before general availability.
Right fit for a small-to-medium team that wants one shared console today and isn’t ready to operate infrastructure for it. Start a Cloud trial.
RetrySight Enterprise — fleet, data residency
Runs entirely inside your network: manager, database, and dashboard on your infrastructure, deployed via Docker or Helm. Four tiers, starting free:
| Tier | Adds |
|---|---|
| Core (free) | Ingest, dashboard, live fleet feed, SSO wizard, Cursor/Claude Code/Codex collectors |
| Analytics | Cost analytics, executive PDF exports, extended collectors, local archive |
| Integration | MCP server, ClickHouse dual-write, Claude Cowork OTel ingest |
| Cloud Scale | Push ingest, object archive (S3/GCS/Azure), git enterprise controls |
Right fit for platform or security teams where “we send agent telemetry to a third-party SaaS” isn’t an acceptable answer. Compare Enterprise tiers.
The rule of thumb
| If you are… | Run… |
|---|---|
| A solo developer evaluating retry visibility | Lite |
| A team of a few developers who want a shared view fast | Cloud |
| A platform team standardizing across 50+ engineers | Enterprise Core, then upgrade tiers as FinOps or MCP needs grow |
Same signals, same taxonomy, different blast radius. Pick based on where the data needs to live today — you’re not locked into that choice tomorrow. Full comparison: Lite vs Cloud vs Enterprise.