Integrations

HYRE Gateway

HYRE LLM Gateway routes Concierge chat through OpenAI-compatible inference — DeepSeek V4 Flash and GLM 4.7 Flash on the same x402 pay-per-call stack. Pick a model in Executive Lounge or pass agentModel on POST /api/concierge.

Status: Live on conc-exe.xyz · Executive Lounge model picker + API

Gateway base: https://gw.hyreagent.fun/api/inference/v1

Partner access: docs.hyreagent.fun/gateway/get-access ↗

Concierge adapter: backend/concierge-api/concierge-hyre.ts

How it works

Operator sets HYRE_GATEWAY_KEY (hyre_gw_*) on the Concierge deployment.
Client selects hyre-deepseek-v4-flash or hyre-glm-4.7-flash in Lounge, or sends agentModel in the chat body.
Concierge maps the id to HYRE catalog models and calls POST /chat/completions on the gateway.
x402 USDC (or Token Pay) settles the Concierge route — consumers do not need a separate HYRE key when calling production.
On gateway error or missing operator key, Concierge falls back to Gemini (modelFallback: true).

Image, enhance, and trading-plan modes always use Gemini. HYRE models apply to mode: "chat" only.

Model map

`agentModel`	HYRE catalog id	Lounge label
`hyre-deepseek-v4-flash`	`deepseek-ai/DeepSeek-V4-Flash`	DeepSeek V4 Flash
`hyre-glm-4.7-flash`	`zai-org/GLM-4.7-Flash`	GLM 4.7 Flash

Aliases accepted in the API: hyre-deepseek, deepseek-v4-flash, hyre-glm. Registry: concierge-llm-models.ts.

Configuration (operators)

Variable	Purpose
`HYRE_GATEWAY_KEY`	Partner gateway key (`hyre_gw_*`) — required for HYRE models on your deploy
`HYRE_GATEWAY_API_KEY`	Alias for `HYRE_GATEWAY_KEY`
`HYRE_GATEWAY_BASE_URL`	Override gateway URL (default `https://gw.hyreagent.fun/api/inference/v1`)

See repo docs/configuration.md and .env.example.

Example API call

pay --sandbox curl https://conc-exe.xyz/api/concierge \
  -H "Content-Type: application/json" \
  -d '{
    "mode": "chat",
    "message": "Summarize BTC vs ETH momentum this week.",
    "history": [],
    "agentModel": "hyre-deepseek-v4-flash"
  }'

Swap agentModel to hyre-glm-4.7-flash for GLM via HYRE. Expect HTML in reply and modelUsed reflecting the upstream model when HYRE succeeds.

Limits

HYRE/GLM chat paths use a tighter LLM timeout so intel prefetch + x402 fit Vercel Edge ~30s.
For structured JSON desks (macro, wire, DeFi intel), use dedicated Intel APIs — faster and model-agnostic.
Default remains gemini when agentModel is omitted.

Concierge Chat API — agentModel and response shape
Poncho — marketplace consumers can request HYRE models on chat route
x402 Protocol
HYRE Gateway access ↗

Open Executive Lounge HYRE Gateway ↗ Chat API