WIR Cloud
Admin Panel
WIR Cloud
Admin
  • Dashboard
  • Billing
  • API Key
  • Pool Accounts
  • Models
  • Combos
  • Playground
  • Users
  • Accounts
  • Proxies
  • Logs
Settings
admin

Dashboard

Token Usage

Last 30 days

Usage by Model

Current billing cycle
ModelProviderRequestsInputOutputCost
No usage yet.

Create Plan

0 means unlimited.

Plans

SlugNameTypePriceTokensReq/dayRPM
Loading...

Usage by User

UserPlanTokens UsedQuotaRequestsErrors
Loading...

Your API Key

••••••••

Regenerate

This will invalidate your current key. All connected clients will need to update.

Generate Pool Accounts (API keys for sale)

Pool Accounts (sold API keys)

UsernameAPI KeyPlanKeysUsage (Tokens)Req / ErrCreatedStatus
No API keys generated yet.

Add Model

Available Models

Model ID Display Name ↑Tier ContextMax TokensPrice InPrice OutAction
No models.

Combos

A combo is one virtual model that tries a list of real models in order. If the first is down, rate-limited, or out of accounts, the next one takes over automatically — one API key, one model name. Buyers you create inherit your combos. Billing follows whichever model actually answers.

letters, numbers, ~ _ . only

    Your Combos

    #NameModels (in order)Action
    No combos yet.

    Create User

    All Users

    #UsernameRolePlanAPI KeyActiveAction
    Loading...

    Add Kiro Account

    Accounts

    #StatusEmailCreditsTokenAddedAction
    No accounts.

    Add XiaomiMiMo Account

    XiaomiMiMo Accounts

    #StatusAPI KeyUsesLast UsedAction
    No accounts.

    Add Hyperagent Account

    Hyperagent Accounts

    #StatusSessionUsesLast UsedAction
    No Hyperagent accounts added yet.

    Add Universal Endpoint

    Endpoints

    #NameBase URLAPI KeyModelsHealthUsesAction
    No endpoints.

    Add Freebuff Account

    Models: fb/deepseek-v4-flash, fb/mimo-v2.5 (any tier) · fb/deepseek-v4-pro, fb/minimax-m3, fb/gpt-5.6-luna, fb/glm-5.2 (need full tier). No base URL needed — WIR-Cloud auto-runs the session + agent-run per request. Hit Test to detect each account's tier.

    Freebuff Accounts

    #StatusAPI KeyTierUsesLast UsedAction
    No accounts.

    Connect Zed Account (login with GitHub)

    Easiest way — no mitmproxy. Click Start, log in with GitHub in the new tab, and when the page tries to redirect to a 127.0.0.1 address that won't load, copy that whole URL from the address bar and paste it below.

    ↗ Re-open sign-in page

    Add Zed Account

    Copy the raw Authorization value from a cloud.zed.dev/client/llm_tokens request (mitmproxy). UID + client id + token are parsed from it; org id is fetched automatically and system id is auto-generated.

    16 models via one Zed client token: zed/claude-sonnet-5, zed/gpt-5.6-sol, zed/gemini-3.1-pro-preview and more. The client token mints a short-lived JWT per request automatically — no re-login. Hit Test to verify a token is live.

    Zed Accounts

    #StatusUIDTokenUsesLast UsedAction
    No accounts.

    Connect ChatGPT Account (login with OpenAI)

    Easiest way — no token scraping. Click Start, log in with your ChatGPT/OpenAI account in the new tab, and when the page tries to redirect to a 127.0.0.1 / localhost address that won't load, copy that whole URL from the address bar and paste it below.

    ↗ Re-open sign-in page

    Add ChatGPT Account

    Bare access token, or Bearer <token>, or Authorization: Bearer <token> — one per line. Prefer the Connect flow above which also saves the refresh token for auto-refresh.

    ChatGPT Plus subscription family: cg/gpt-5.6-sol, cg/gpt-5.4, cg/gpt-5.5 and more — routed via the Codex Responses API (no OpenAI API billing). Requires a ChatGPT Plus (or higher) account per token. Hit Test to verify a token is live.

    ChatGPT Accounts

    #StatusEmailTokenUsesLast UsedAction
    No accounts.

    Connect Antigravity Account (login with Google)

    Easiest way — no token scraping. Click Start, log in with your Google account in the new tab, and when the page tries to redirect to a 127.0.0.1 / localhost address that won't load, copy that whole URL from the address bar and paste it below. Project bootstrapping + Gemini Code Assist onboarding run automatically.

    ↗ Re-open sign-in page

    Add Antigravity Account

    Bare access token, or Bearer <token>. Prefer the Connect flow above which also saves refresh token + auto-bootstraps the project id / tier.

    Gemini Code Assist family: ag/gemini-3.7-flash-high, ag/gemini-pro-agent, ag/claude-sonnet-4-6 and more via the Antigravity IDE client. Requires a Google account enrolled in Gemini Code Assist. Hit Test to verify a token is live.

    Antigravity Accounts

    #StatusEmailProjectTokenUsesLast UsedAction
    No accounts.

    Add Proxies

    Zed blocks datacenter IPs for trial accounts (trial_blocked) by region, so Zed calls egress through this pool. Hit Vet after pasting: it geo-checks each proxy's country and fires a live Zed probe, marks blocked IPs dead, and flags the survivors ✓ vetted. The router always prefers vetted proxies, so live requests never wander into a blocked IP. A blocked IP is kept for history (not auto-deleted); re-vet anytime.

    Freebuff Auto-Proxy

    Egress Freebuff through the proxy pool
    Auto-proxy Freebuff via Freebuff proxy pool
    ON: Freebuff requests egress through proxies added above with provider Freebuff — one IP holds many accounts until it hits Freebuff's per-IP session cap (ip_capped), then requests roll to the next proxy automatically. If every proxy is exhausted/dead, it falls back to the direct VPS IP (risk of per-IP cooldown). OFF: always use the direct VPS IP. Global admin switch.

    Proxy Pool

    #StatusProxyCountryOKFailLast UsedAction
    No proxies.
    Playground · ephemeral, not saved
    ⚙ Settings
    Pick a model, type a message. Each turn is an independent test — history is kept only in this view and cleared on refresh.

    Request Logs

    TimeProviderModelAI AccountStatusTokensUserDuration
    No logs.

    Compression

    Reduce token usage for long conversations
    RTK Conversation Compression
    Auto-compress long conversations by keeping the first and last messages and summarizing the middle.
    Caveman Terse Responses
    Injects a system prompt that makes the model reply more concisely.

    Vision Captioner

    Auto-describe images for non-vision models
    Auto-caption Image fallback for non-vision models
    When a non-vision model receives images, the chosen captioner model describes them and the descriptions are passed instead. Lets text-only models handle images.
    Must be a model that supports image input. Validation is performed before save.

    Change Password

    Allowed Models —

    Select which models this user can access. Empty selection means all models are allowed (default).

    Add API Key

    Add one or more API keys to this endpoint. Enter one key per line for bulk add.