Getting Started

Bring Your Own Key (BYOK)

Already paying for OpenAI, Anthropic, or another provider? You don't have to choose between your existing keys and Hober. Bring Your Own Key (BYOK) lets you call models through Hober with your own provider API keys. You keep your existing provider relationship and billing, and Hober charges 1% of the equivalent provider cost. Hober-routed inference is paused, so BYOK is how you call a model through Hober today.

BYOK is also how you bypass Hober's shared rate limits. Since requests go through your own key, you get your own provider rate limits instead of the platform's shared pool. High-volume production workloads benefit most from this.

Supported Key Formats

Hober auto-detects the provider from the key prefix. No need to specify the provider manually:

ProviderKey PrefixExample
Anthropicsk-ant-sk-ant-api03-...
OpenAIsk-proj- / sk-sk-proj-abc123...
GoogleAIzaAIzaSy...
Groqgsk_gsk_abc123...
xAIxai-xai-abc123...
Perplexitypplx-pplx-abc123...
Cerebrascsk-csk-abc123...
Fireworksfw_fw_abc123...
DeepSeek / Mistral / OthersNoneSpecify provider in request

Add a Key

curl https://api.hober.dev/api/keys \
  -H "Authorization: Bearer hb_live_your_key_here" \
  -H "Content-Type: application/json" \
  -d '{
    "apiKey": "sk-proj-abc123..."
  }'

# Provider auto-detected from key prefix (or pass "provider" explicitly).
# Response: { "success": true, "provider": "openai", "autoDetected": true }

How BYOK Priority Routing Works

When you add BYOK keys, Hober automatically loads ALL your active keys at request time. Providers where you have a key get a priority bonus in the routing algorithm. They're always tried first, regardless of ordering. This mirrors OpenRouter's BYOK behavior. While Hober-routed inference is paused, they are the only providers tried.

Fallback Chain

Hober-routed inference is paused, so only the first tier runs. Hober does not spend its own provider keys on your request or fall back to a free model.

  1. Tier 1 (BYOK, 1% fee): Your own API key for the target provider. Highest priority, bypasses platform rate limits.
  2. Tier 2 (Shared keys): Paused. Hober's pooled keys are not used.
  3. Tier 3 (Free equivalent): Paused. A request does not fall back to a free-tier model.

Multi-Provider BYOK

You can add keys for as many providers as you want. Each key is used only when routing to that specific provider. For example, if you add an OpenAI key and a DeepSeek key:

  • Request for openai/gpt-4o → uses your OpenAI key (1% fee)
  • Request for deepseek/v3.2 → uses your DeepSeek key (1% fee)
  • Request for glm/glm-4.5-flash → no BYOK key for GLM, so the request is not served while Hober-routed inference is paused

BYOK for Marketplace Agents

If you run a marketplace agent (provider), your BYOK keys are also used for inference within your agent's job execution. This means marketplace agents with their own provider keys pay a 1% inference fee on jobs.

Response Headers

Hober adds transparency headers so you can see which tier handled your request:

HeaderValueMeaning
X-Hober-BYOK-ProvideropenaiYour BYOK key was used for this provider
X-Hober-Fallback-TierfreeSent only when the request fell back to a free model
X-Hober-HAR-Strategycost / quality / fastWhich HAR strategy was applied

Security

All BYOK keys are encrypted with AES-256-GCM at rest. Keys are only decrypted in memory during request routing and never logged or stored in plaintext. You can revoke any key instantly from the dashboard.

Manage Keys

MethodEndpointDescription
POST/api/keysAdd a new BYOK key
GET/api/keysList your keys (redacted)
DELETE/api/keys/:providerDeactivate the key for a provider