Can you self-host Raindrop?

What Raindrop Offers Today

Raindrop is a hosted service. Its 2.0 announcement, on 3 June 2026, says: "The whole loop can run in your own cloud. Raindrop 2.0 now deploys directly in your VPC, so traces and agent data never leave your infrastructure." Two sentences later: "We're rolling VPC out with a select group of initial partners." The homepage says the same thing more briefly - "Run Raindrop in your own cloud when your deployment requires it" - and marks it beta.

The VPC capability exists, but it is not generally available. Raindrop publishes no self-hosting guide, system requirements or deployment chart. Teams that need it must contact Raindrop about joining the partner rollout. No on-premises or air-gapped option is listed.

For many security reviews, the cloud is acceptable with the right controls, and Raindrop has several: SOC 2 Type II, PII redaction, and - on Enterprise - SSO and SAML, audit logs, edge PII redaction and warehouse export.

What You Can Run Yourself Today

• Workshop. An MIT-licensed local debugger and eval runner, launched in May 2026. It runs on a developer machine and is useful for debugging and evaluating an agent locally. It is not the production platform: no Signals across your traffic, no Experiments.
• Client-side redaction. The TypeScript SDK can redact with regex patterns before an event leaves your process. It is TypeScript-only, and regex will miss what it was not written for.
• Choosing what you send. Every event is sent by your code, so you decide which turns and which fields go. Sending less reduces exposure and reduces what Raindrop can detect in equal measure.

Note what PII Guard does and does not do here: it runs server-side at ingestion, so the data has already reached Raindrop when it is redacted. The Enterprise "edge PII redaction" is the option aimed at redacting earlier; ask how and where it runs.

If You Need Self-Hosting Now

If keeping data in your own cloud is a hard requirement, these are the available options and what each asks your team to operate.

The free options avoid a platform licence fee, but they still require backups, upgrades, capacity planning and incident response. Latitude provides self-hosted tracing, evaluations and Agent Dispatch using the same images as its cloud. None of the options in the table has a direct equivalent to Raindrop Experiments for comparing cohorts of logged production traffic.

Failproof AI Enterprise is another option when you need agent observability in your VPC before Raindrop's rollout is generally available. The deployment keeps traces and evaluation data in your environment and includes querying, code-based and LLM-based evaluations, automated failure analysis, findings, alerts and policy management.

What to Ask Raindrop About the VPC Rollout

• Which components run in our VPC, and does any data or telemetry still go to Raindrop's cloud?
• Who operates it - Raindrop staff with access to our account, or our team with images you ship?
• How are upgrades delivered, and how far behind the hosted product can a VPC deployment run?
• Which clouds and regions are supported, and what does it cost relative to Pro and Enterprise?
• What are the criteria for joining the partner group, and is there a date for general availability?

When You Do Not Need to Self-Host

If your events contain no regulated data, or you can keep it out with client-side redaction and careful choice of fields, the hosted service is simpler and cheaper to run than any self-hosted stack on this page. Self-hosting an observability platform to satisfy a preference rather than a requirement usually costs more in engineering time than it saves in risk. Ask your security team which of the two it is before you plan the cluster.