Skip to content

Contrast vs Elastic Observability

Elastic Observability stores each span as its own document, and its Serverless plans bill every GB you ingest. Contrast keeps each run together from the moment it lands. Questions about what happened next read straight off the run.

Updated November 2026

At a glance

7.8×
faster than ClickHouse
90%
lower cost than Snowflake
$0
monthly fee

Recorded races, October 2026. Elastic Observability wasn't in the test.

How you're billed

List prices as each vendor publishes them, checked November 2026.

Contrast

Billed on
Storage, data scanned and AI credits
Price
$10 per TB scanned

No monthly fee on pay as you go, and the Free plan doesn't expire.

Elastic Observability

Billed on
Data ingested and retained
Price
From $0.09 per GB ingested

Plus $0.019 per GB retained each month, on Serverless Complete.

Rebuilding agent runs out of span documents?

Agent questions are about order: what a run did after an error, and what it tried next.

Contrast

TraceDB stores each trace as one ordered timeline. Its indexes point at traces, so a question about a rare event reads only the runs that contain it.

Elastic Observability

Telemetry lands as documents in data streams, one document per span.

What agents did after a tool error

TraceDB returns the paths themselves, counted across every run. The ES|QL query returns every span after the error and leaves the counting to you or your agent.

Contrast
-- the two steps after each tool error
SELECT path, count, entities
FROM paths(from => 'tool_error', steps => 2)
ORDER BY count DESC
LIMIT 10
Elastic Observability
FROM traces-*.otel-*
| INLINE STATS error_at = MIN(@timestamp)
    WHERE span.name LIKE "execute_tool *"
      AND status.code == "Error" BY trace_id
| WHERE @timestamp > error_at
| SORT trace_id, @timestamp
| KEEP trace_id, @timestamp, span.name
| LIMIT 1000

Contrast and Elastic Observability, feature by feature

ContrastElastic Observability

Runs and queries

Each trace kept as one ordered runContrastYes. Spans stored together, in time order.Elastic ObservabilityPartly. Separate documents that share a trace ID.
Functions for what came nextContrastYes. paths(), followed_by() and preceded_by().Elastic ObservabilityPartly. No next-step function.
SQL over tracesContrastYes. Standard SQL plus trace functions.Elastic ObservabilityPartly. ES|QL, a piped language of its own.

Agent runs

Model and tool calls named in each runContrastYes. With status, tokens and model.Elastic ObservabilityPartly. GenAI attributes on each span document.
Reads the common agent tracing conventionsContrastYes. GenAI, AI SDK, OpenInference, Langfuse and more.Elastic ObservabilityPartly. Stored as sent under attributes.*.
Drop unmapped attributes from the exporterContrastYes. One header: X-Contrast-Trace-Unmapped: drop.Elastic ObservabilityPartly. Collector processors you configure.

Agents

Hosted MCP serverContrastYes. OAuth at contrastdb.com/api/mcp.Elastic ObservabilityYes. Agent Builder MCP server, a paid add-on.
Trace tools for agentsContrastYes. engine_describe, engine_query and engine_sql.Elastic ObservabilityPartly. Agent Builder tools like an ES|QL query.
Traces and the warehouse in one workspaceContrastYes. Iceberg warehouse beside TraceDB.Elastic ObservabilityPartly. Cross-project search is an add-on.

Storage and pricing

No data tiers to manageContrastYes. S3, cached on disk and in memory.Elastic ObservabilityPartly. Lifecycle tiers or a Search Boost Window.
No charge per GB ingestedContrastYes. Storage, data scanned and AI credits.Elastic ObservabilityNo. From $0.09 per GB ingested.
No monthly feeContrastYes. Pay as you go, after a Free plan.Elastic ObservabilityYes. Serverless bills usage only.

Running alongside Elastic Observability

  1. A workspace owner or admin starts TraceDB from Sources, under Agent traces. It takes a few minutes.
  2. Create an API key on the same page. Contrast shows it once, with the OTEL_EXPORTER_OTLP_TRACES variables your SDK reads.
  3. Add an otlphttp exporter for https://api.embrasure.ai/v1/traces to your OpenTelemetry Collector, next to the one that sends to Elastic, so both get the same spans.
  4. Name each run, then ask about runs in Contrast or through its MCP server.

More comparisons

Speed figures are from TraceDB races recorded October 27, 2026 on a six-query investigation: TraceDB and each rival on 4 vCPUs of one 8-vCPU VM in AWS us-east-1, each run alone, the median of 5 races after a warm-up, over 100,137,865 events. Elastic Observability was not part of that test. The cost figure is the one the Contrast home page states for TraceDB against Snowflake. Elastic Observability details are from its public documentation, checked November 2026: Elastic Observability Serverless pricing, OpenTelemetry with Elastic APM, OpenTelemetry data streams, ES|QL reference, ES|QL INLINE STATS, Data tiers, Serverless project settings, Agent Builder MCP server. Product names are trademarks of their owners.