Provider key onboarding probe

Portkey AI Gateway | key inventory fit

Portkey AI Gateway key inventory fit decides whether a key can move from identification into inventory, pricing, routing, and launch. TestKey starts with 2 covered models, global route, sample models gpt-4o-mini · claude-3-5-haiku-latest, then verifies with read-only probes.

Provider

Portkey AI Gateway

global route

Model coverage

2

gpt-4o-mini · claude-3-5-haiku-latest

Probe dimension

key inventory fit

inventory-fit

Onboarding signal

2 models can become inventory candidates

portkey

Read-only probes

1

current_key

Sample metrics

7

key_valid / requests_total / runtime_model / runtime_model_allowed

Ten capabilities

4

operational actions / model permission matrix / key identity

Provider onboarding summary

Provider

Portkey AI Gateway

Probe dimension

inventory-fit

Model coverage

2

Read-only check. Detection data burns after 5 minutes.

Why this probe matters

Portkey AI Gateway key inventory fit decides whether a key can move from identification into inventory, pricing, routing, and launch. TestKey starts with 2 covered models, global route, sample models gpt-4o-mini · claude-3-5-haiku-latest, then verifies with read-only probes.

Provider: Portkey AI Gateway
Model coverage: 2
Probe dimension: key inventory fit

How to prove it

Read-only check. Detection data burns after 5 minutes.

Identify the real provider first, then verify key inventory fit; do not treat OpenAI-compatible as real OpenAI by default.
Use 2 covered models as candidates, then read /models, balance, error bodies, and limit headers.
2 models can become inventory candidates

Next action

Portkey AI Gateway key inventory fit should become an action: add provider config, fill pricing, import inventory, set route, add monitoring, or send to human review.

Read-only check. Detection data burns after 5 minutes.
Onboarding signal: 2 models can become inventory candidates
Model coverage: gpt-4o-mini · claude-3-5-haiku-latest

Ten key capability anchors

Portkey AI Gateway has 1 read-only account probes configured, currently supporting operational actions / model permission matrix / key identity / capability matrix. These anchors feed the real TestKey detection path, not just admin display.

Read-only probes: current_key, used to prove provider identity, models, balance, or subscription state.
Sampleable metrics: key_valid / requests_total / runtime_model / runtime_model_allowed / token_type / token_role / token_scopes.
Ten-capability anchors: operational actions / model permission matrix / key identity / capability matrix.

Open key checker

Read-only check. Detection data burns after 5 minutes.

Open providers

Portkey AI Gateway

Open models

gpt-4o-mini · claude-3-5-haiku-latest