Observability for every request

Latency, tokens and cost per request, broken down by model, router and customer.

Start for Free
Acme Inc · org_3kf9d2

Overview · Traffic

Customize Save view
Time:
1h 24h 7d 30d
Custom Filters 2
Export
Requests
128,403
Success rate
99.4%
5xx errors
142
Tokens
84.2M
Latency · p95
1,240ms
Latency · p99
3,180ms
Cost
$1,284.50
Customers
312
Requests · Count / time 128,403
Outcome · 5xx / time 142
Latency · p95 / time 1,240ms
Cost · Total / time $1,284.50
Requests 8 loaded+
Columns
TimeRequest IDRouterAliasModelCustomerIn/OutCostLatAttStatus
07:32:14.291 req_8f2a91 rt_chatbot smart claude-sonnet-4-6 tenant_acme 1,204 / 856 $0.0142 842ms 1 200
07:32:13.884 req_7d10c4 rt_agent — gpt-5 user_42891 2,860 / 1,420 $0.0388 1,910ms 1 200
07:32:12.530 req_5c93af rt_summarize fast claude-haiku-4-5 tenant_globex 640 / 220 $0.0021 318ms 1 200
07:32:10.119 req_4b77e2 rt_agent smart gpt-5 tenant_initech 3,120 / 980 $0.0402 2,240ms 2 200
07:32:08.742 req_3a18bd rt_chatbot cheap gemini-2.5-pro user_10233 1,540 / 612 $0.0096 1,120ms 1 200
07:32:06.501 req_29ee70 rt_embed — text-embedding-3-large tenant_acme 820 / 0 $0.0003 92ms 1 200
07:32:04.330 req_1f0c55 rt_agent smart gpt-5 tenant_hooli — — — 3 429
07:32:01.987 req_0b6d18 rt_chatbot fast claude-sonnet-4-6 user_88120 — — 1,480ms 2 503
8 rows Load more

Saved views

Built-in views that can be edited and extended.

Requests
1.24M
Success
99.6%

Overview

Request volume, success rate and spend.

p95
1.84s
TTFT p50
240ms

Latency

p50 to p99, time to first token.

Error rate
0.4%
Recovered
92%

Errors

Rate limits, timeouts and provider errors.

Spend
$4,812
Per request
$0.0039

Cost

Spend, cost per request and per token.

Models
9
Top
Sonnet 4.6

By model

Traffic share by model.

Customers
1,208
Top
tenant_acme

Customers

Usage and cost for each of your customers.

Request detail

The route, latency and cost of each request.

Request req_7Kd2Qx9WmB4Tn6Ls200
Latency
812 ms
Cost
$0.0210
Tokens
12,800
TTFT
240 ms
Attempts
2
Router overhead
6 ms
Routing
Model servedanthropic/claude-sonnet-4.6
Vendor · regionbedrock · us-west-2
Routerrt_agent
Customertenant_acme
Where the time went
UploadRouterUpstreamResponse
Cost
Input 2,000 tokens$0.0060
Cache read 10,000 tokens$0.0030
Output 800 tokens$0.0120
Total$0.0210

Priced from the offering that served it, and stored with the request.

End-to-end tracing

Requests are tagged on ingress and searchable by tag.

You sendPOST /v1/chat/completions
x-mr-session-id: conv_8812
x-mr-correlation-id: job-2026-10-01-0042

{
  "model": "multi-router/smart",
  "user": "tenant_acme",
  "messages": [ … ]
}
You get back200 OK
x-mr-request-id: req_8Fq2LmXo1c
x-mr-correlation-id: job-2026-10-01-0042
x-mr-provider: bedrock
x-mr-region: us-west-2
x-mr-upstream-model: anthropic/claude-sonnet-4.6
x-mr-attempts: 2
x-mr-composer: rt_agent

Request logging

Configure what is stored for each request.

Request metadata
FullAnonymousFailures onlyNone
Payloads
AllSample 10%Failures onlyNone

Per router

Keep full records, drop customer and session ids, or store failures only.

Traffic logging
Anonymous metadata (usage only)
All request metadata
Store everything

Storage Policy

Administrators set the ceiling for every router in the organization.

Authorization[redacted]
x-api-key[redacted]
X-MR-Api-Token[redacted]
cookie[redacted]
Encrypted, stored per organization

Protected payloads

Credentials are stripped before anything is stored.

Filter, save and export

Model: claude-sonnet-4.6Status: 5xxCustomer: tenant_acmeLast 7 days
Search request, session or correlation id

Filter anything

Model, router, provider, status, error kind, customer and more.

Checkout errors
EU latencyModified
Weekly cost

Saved views

Save filters, time range, statistics and charts as a view.

traffic-2026-10-13.csvCSV
Rowsup to 100,000

Export

Download any view as CSV.

Send a request and open it in Traffic

Each request appears in Traffic with its latency, tokens and cost.

Start for Free