Tag at request time
Pass a request_tags array with any chat completion and it's tracked at the individual request level-no new API keys and no dashboard setup.
Dynamic Tags
Tag every LLM request, analyze usage your way
Add a request_tags array to any chat completion and FastRouter organizes usage by feature, customer, environment, team, or geography-filterable in the dashboard and captured in your activity logs, with no extra setup.
No credit card required · Free to start
Request inspector
POST /v1/chat/completions
200 OKmodelopenai/gpt-4o-mini
request_tags
feature/summary-widgetcustomer/acmeenv/production
Usage by tag
last 7 days- feature/summary-widget46%
- customer/acme32%
- env/production22%
Why Dynamic Tags
Usage attribution without the overhead
Label requests the moment you send them, then slice cost and usage by any dimension that matters-no new keys, no pipelines to maintain.
Flat or hierarchical
Use simple one-level tags like production, path-style hierarchies like team/feature/test, or mix both formats in the same request.
Filter & analyze anywhere
Tags become a dashboard filter and land in activity-log metadata, so you can slice usage and cost by any dimension you choose.
How it works
From tagged request to clear analytics
Tag a request, route it as usual, and FastRouter records every tag-ready to filter and report on across the dashboard and your activity logs.
Request
Tagged chat completion
request_tagsflathierarchical
- Send any chat completion with a request_tags array.
- Mix flat and path-style tags in the same call.
FastRouter
Gateway
Routes & metersRecords tagsNo extra setup
- Routes the request and records every tag it carries.
- Works with your existing API key-no new keys to manage.
Logs & analytics
Filter & report
Dashboard filterActivity logsUsage reports
- Tags become a filter in the dashboard.
- Stored in activity-log metadata for reporting.
Per-request tagging
Set tags inline on every request
request_tags is a simple array of strings you pass with any chat completion. There's nothing to provision-tag a call and it's instantly tracked at the request level.
One array, zero setup
Add a request_tags array of strings to any request-no new keys, no configuration, no pipelines to maintain.
Drops into your stack
request_tags is just another field on the OpenAI-compatible request body, so it works with your existing SDKs and keys.
Flat, nested, or mixed
Send simple tags like production or path-style tags like org/team/project-and combine them in the same call.
POST /api/v1/chat/completions
JSON{
"model": "openai/gpt-4o-mini",
"stream": false,
"messages": [{ … }],
"temperature": 0.0,
"request_tags": ["UK/London", "Europe"]
}
Tags recorded with the requestFilterable in dashboard
Flat & hierarchical
Flat tags or nested hierarchies
Choose the structure that fits your reporting. Flat tags group quickly, while path-style tags let you roll usage up by any level of the path-and you can mix both in one request.
Flat tags
Simple one-level labels such as asia or production for fast, lightweight grouping.
Hierarchical paths
Path-style tags like team/feature/test roll usage up by any level of the hierarchy.
Mix both freely
Combine flat and nested tags in the same request_tags array-no rigid schema required.
Tag taxonomy
Mix & matchFlat tags
asiaproductionbeta
Hierarchical paths
- customer/acme
- org/team/project
- US/California/SF
Filter & analyze
Filter and break down usage by tag
Tags surface as a dashboard filter and live in activity-log metadata, so you can isolate a feature, customer, or environment and attribute spend without juggling extra API keys.
Dashboard filter
Filter requests by tag right from the dashboard to isolate a single feature, customer, or environment.
Activity-log metadata
Every tag is stored in the metadata of your activity logs for granular tracking and auditing.
Usage reports
Organize cost and usage reports by tag-without provisioning and managing multiple API keys.
Activity logs
1 filterrequest_tagscontainscustomer/acme
- 09:41$0.42
gpt-4o-mini
customer/acmefeature/chat - 09:40$1.08
claude-3.5-haiku
customer/acmeenv/production - 09:38$0.31
gpt-4o-mini
customer/acmefeature/summary
128 requests matched$54.20 attributed
Tag formats
Pick the tag shape that fits your reporting
Flat, hierarchical, or both-every format travels in the same request_tags array and shows up the same way in filtering and logs.
| Capability | Flatasia | Hierarchicalorg/team | Mixedboth |
|---|---|---|---|
| Tag structure | |||
| Simple one-level tags | Included | Not included | Included |
| Path-style nested tags | Not included | Included | Included |
| Flat and nested in one request | Not included | Not included | Included |
| Organize & analyze | |||
| Filter requests in the dashboard | Included | Included | Included |
| Roll-up reporting by tag path | Not included | Included | Included |
| Stored in activity-log metadata | Included | Included | Included |
All formats are passed in the same request_tags array-no extra configuration required.
Use cases
One tag, many ways to slice usage
From per-customer billing to feature-level cost tracking, dynamic tags turn raw request volume into the breakdowns your teams actually need.
Feature usage breakdown
Tag each feature like feature/summary-widget to see which ones drive the most tokens, latency, and spend.
Customer cost attribution
Attribute usage to customers or teams with tags like customer/acme for chargebacks and per-customer billing.
Environment segmentation
Separate production, staging, and experiments with tags like env/staging-no new API keys required.
Geo & team tracking
Break usage down by region or team using path-style tags like US/California/SF or marketing/email-gen.
FAQ
Dynamic tags, answered
Dynamic tags are custom labels you attach to individual requests with the request_tags parameter-an array of strings on the chat completions request. They enable tracking at the request level, so you can organize, filter, and analyze usage without changing how you call the API.
Include a request_tags array of strings in your chat completions request body, for example request_tags: ["UK/London", "Europe"]. Tags are set per request and work with your existing API key, so there's no extra setup or configuration.
Three formats: flat tags like ["asia", "production"], hierarchical path-style tags like ["USA/NYC", "org/team/project"], and a mix of both in the same request. Hierarchical paths let you roll usage up by any level of the path.
Tags are available in the dashboard as a filter and are stored in the metadata of your activity logs. That lets you group and filter requests by project, team, environment, or feature and build usage reports around them.
No. Dynamic tags let you track granular usage-by team, feature, customer, environment, or geography-without provisioning and managing multiple API keys. A single key plus tags is enough to attribute usage.
Common uses include cost allocation and team billing (customer/acme, env/staging), feature usage breakdowns (feature/summary-widget), team-level visibility (marketing/email-gen), and geo-based tracking (US/California/SF). Any dimension you can express as a tag becomes a way to slice your usage.
Start tagging your LLM usage today
Add request_tags to your next call and watch usage organize itself by feature, customer, environment, and team-no extra setup.