Talk to an engineer

Telemetry Policies

Safe production change for telemetry data.

Give teams and AI agents a safe way to change telemetry: Telemetry Policies that drop, redact, sample, or route data before storage.

Talk to an engineerRead the spec
Clean signal
host · ip-10-0-12-44 · us-east-1
OTel Collector
Tero policy engineGo
Policies synced

The problem

Telemetry change was not built for the AI era.

Collector configs, filters, transforms, scripts, and code changes work when humans make occasional edits. They break when teams and AI agents can find thousands of precise telemetry fixes.

AI-era production
More AI agents
More code
More features
More deploys
More services
More logs
Manual telemetry controls
Collector configsTransform rulesExclusion filtersSampling rulesOne-off scriptsCode changes
Operational drag
Bill spikes
Sensitive data in logs
Renewal shock
Backlogged fixes
Risky deploys
Unclear ownership

The old model turns every fix into platform work.

Telemetry Policies change that.

Primitive

One policy, one intent.

A Telemetry Policy is a small YAML file that makes one telemetry change: drop, redact, sample, route, or reshape data before storage.

  • File-backed
  • Atomic
  • Scoped
  • Reviewable
  • Portable
  • Reversible

Built on an open standard, donated to OpenTelemetry as OTEP-4738 ->

Traceability

Every policy has a lineage.

Policies stay linked to the log pattern, issue, owner, and verification state, so teams can see why a change exists and what it fixed.

  1. Log pattern foundmiddleware_request
  2. Issue openedISS-1247
  3. Policy createdPOL-1042
  4. Fix verifiedmatches drop

Runtime

Runs where telemetry flows.

Policies deploy into OpenTelemetry Collector, Vector, Datadog Agent, Tero Edge, and embedded engines.

Clean signal
host · ip-10-0-12-44 · us-east-1
OTel Collector
Tero policy engineGo
Policies synced
OTel dist·OTel guide

Scale

Built for thousands of changes.

Instead of piling rules into fragile pipeline configs, policies compile into efficient runtime behavior built for AI-scale fix volume.

Benchmarked throughputView benchmark data ->
Requests / second
Number of policies
Policy engineOTel Collector rulesVector transforms

Ownership

Ownership stops being platform work.

Platform keeps the guardrails. Service teams own the telemetry changes tied to their services.

@checkout
@api-gateway
@auth
@ai-agent
platform guardrailsschemalimitsreviewapproval
OTel Collector
Vector
Datadog Agent
Tero Edge

Give teams and agents a safe way to change telemetry.

Use Telemetry Policies to turn log issues into small, reviewable changes that run where your telemetry already flows.

Talk to an engineerRead the specView benchmarks