Observability forces an expensive decision: keep everything hot, or filter, sample, and shorten retention. The value of a log often becomes clear only after an incident—sometimes weeks after you decided whether to keep it.
Ryo separates retention from hot indexing. Keep Datadog, Splunk, or New Relic. Retain the full-fidelity copy in storage you own, send selected telemetry to your existing platform, and search the rest when you need it.
Architecture
Your observability platform becomes a materialized view over your telemetry. It no longer needs to be the system of record.
Ryo Gateway receives telemetry before the destination vendor, writes the full copy, and applies explicit routing policy. Existing dashboards, monitors, APM, and incident workflows stay in place.
Ryo Vault is the durable canonical copy in customer-owned storage. Keeping more history should not require a growing always-on search cluster: retention cost should follow storage, and query compute should follow demand.
Ryo Search makes the retained copy operationally useful. Investigate by service, time range, request or trace ID, namespace, container, structured fields, or text—including data never sent to the hot tier.
Agent access
Connect your own agents through MCP or an API. Let them follow requests across services, search older events, and test explanations against the full history on your behalf.
AI sovereignty means you choose the model, the agent, and the data it can access. Give it room to follow leads within those permissions, rather than limiting investigations to whatever survived hot-tier retention.
Ryo’s pricing principle is no separate fee per MCP or API call. Storage, query compute, and your model provider’s charges still apply.
How other vendors meter access
New Relic meters Core Compute Capacity Units for actions including queries and API calls.
Elastic lists Agent Builder executions from $0.025 after 1,000 free executions. This is an agent execution charge, not a blanket MCP fee.
Axiom counts agent queries toward normal query-hour usage, with no separate MCP pricing. A per-call surcharge is not universal.
Sources checked September 5, 2026. These meters cover different services; they are not a total-cost comparison.
Usage-aware placement
Ryo Advisor compares source volume with query activity and monitor and dashboard dependencies to identify candidates for less expensive retention.
| Source | checkout INFO |
|---|---|
| Monthly volume | 8.4 TB |
| Queries in 30 days | 17 |
| Known monitor dependencies | None |
| Known dashboard dependencies | None |
The long-term goal is automatic placement based on use, latency, cost, and retention requirements. Low observed use is a reason to reconsider the tier, not discard the evidence.