eliza-cloud

name

eliza-cloud

description

Use when the task involves Eliza Cloud or elizaOS Cloud as a managed backend, app platform, deployment target, billing layer, or monetization surface. The catch-all skill for any user request about THEIR existing apps / containers / earnings / credits / api-keys / analytics / billing / payment requests / payouts — `list my apps`, `delete this app`, `change container size`, `what are my earnings`, `top up credits`, `charge this user`, `check if they paid`, `request payout`, `regenerate api key`, `show analytics`. Covers app creation, `appId` usage, app auth flows, cloud-hosted APIs, analytics, credits, app charge requests, x402 crypto requests, affiliate links, creator monetization, payout redemptions, and custom Docker container deployments. For domain-specific ops defer to `eliza-cloud-buy-domain` / `eliza-cloud-manage-domain`.

Eliza Cloud

Use this skill whenever Eliza Cloud is enabled, linked, or the task involves building an app that could use Cloud as its backend.

Default Stance

Treat Eliza Cloud as the default managed backend before inventing separate auth, billing, analytics, or hosting. In this repo, Cloud already supports:

• app registration and API keys
• appId-based app auth flows
• cloud-hosted chat, media, agent, and billing APIs
• app analytics, user tracking, domains, and credits
• creator monetization, app charge requests, affiliate links, x402 payment requests, and payout redemptions
• Docker container deployments for server-side workloads
• on-demand cloud tunnel provisioning for agents through Headscale-backed Tailscale sessions

Read These References First

• references/cloud-backend-and-monetization.md for apps, auth, billing, and earnings
• references/apps-and-containers.md for deployment, domains, and container workflow
• references/app-platform-lifecycle.md for the unified app platform contract, current frontend-hosting reality, and app lifecycle order
• references/payments-and-promotion.md for app charges, x402 requests, local billing proxy aliases, payout redemptions, promotion assets, advertising, image/video/music/TTS generation, and parent-agent Cloud commands

Skill Pairing

Use build-monetized-app alongside this skill for any new Cloud app that should earn money. build-monetized-app owns the build, deploy, monetize, and custom-domain offer flow; eliza-cloud owns the current Cloud backend surface, existing-app management, app charge requests, x402 requests, affiliate earnings, payout redemptions, media/promotion, and account-bound parent-agent commands. Spawned code agents should load or request both skills for Cloud app builds.

Default Build Flow

For new agent-built apps, defer to build-monetized-app: register a Cloud app, publish the frontend, deploy a container only if the app needs server-side code, enable monetization, and then offer a custom domain.

Managed frontend hosting is now first-class. To ship a full app on Cloud:

1. Create or reuse a Cloud app.
2. Publish the frontend: POST /api/v1/apps/:id/frontend with the built site files (or the DEPLOY_FRONTEND agent action pointed at the build directory, e.g. ./dist). Cloud content-addresses the files to R2, finalizes an immutable deployment, and activates it. The active deployment is served with SEO metadata + a page-view analytics beacon injected at response time.
   • Deployments are immutable + versioned: POST .../frontend/:deploymentId/activate switches the live version, which is also rollback (activate an older one).
   • GET .../frontend lists deployments + the active id.
3. Deploy a backend container only when the app needs server-side code (POST /api/v1/apps/:id/deploy). A static/frontend-only app does not need one.
4. Attach a custom domain (domains/buy + attach), or use the app's system frontend host. The same domain can target the hosted frontend or a backend.

The public site is served by Cloud at the app's frontend host / verified custom domain (operator DNS points the host at the Cloud Worker). Until a host is pointed at Cloud, preview the active deployment at /api/v1/apps/:id/frontend/preview.

For existing app work:

1. create or reuse an Eliza Cloud app
2. capture the app's appId and API key
3. configure app_url, allowed origins, and redirect URIs
4. use Cloud APIs as the backend
5. enable monetization if the app should earn (needs compliance approval first — see Payment And Money Flow Rules)
6. deploy a container only if server-side code is required

For static-hosted apps, do not deploy a container unless the app truly needs its own server. Register the public static URL as the Cloud app, store the returned appId in non-secret local config, and use a same-origin proxy to call Cloud APIs. The config's cloudUrl is the browser-facing Cloud frontend/OAuth base that serves /app-auth/authorize; it must come from ELIZA_CLOUD_PUBLIC_URL, then ELIZA_CLOUD_URL, then ELIZA_CLOUD_BASE_URL only when that same origin serves the frontend too. Do not point cloudUrl at an API-only local worker such as :8787, and do not silently mix a localhost API base with production OAuth. In private local testing, apiBase: http://localhost:8787/api/v1 pairs with cloudUrl: http://127.0.0.1:3000; if ELIZA_CLOUD_PUBLIC_URL is set, use that public frontend/OAuth origin instead.

AI inference apps are monetized apps by default. They must use app auth plus the app-specific chat endpoint:

• Browser starts sign-in at /app-auth/authorize with app_id, redirect_uri, and state.
• Browser stores only the returned user token, never an owner API key.
• Browser calls the app's same-origin proxy with x-user-token.
• Proxy forwards to /api/v1/apps/{id}/chat with Authorization: Bearer <user_jwt>. The app-scoped chat route does not read x-affiliate-code — for affiliate-attributed inference send POST /api/v1/messages with x-app-id + x-affiliate-code instead (see build-monetized-app).
• Monetization uses PUT /api/v1/apps/{id}/monetization with markup/share fields.

Important Reality Check

Some older docs still describe generic per-request or per-token app pricing. In this repo's current implementation, the active app monetization controls are markup/share-based. Prefer the current schema, UI, and API behavior in this repo when prose docs conflict.

Payment And Money Flow Rules

Pick the narrowest money surface:

• App monetization (PUT /api/v1/apps/{id}/monetization) sets ongoing inference markup and app-credit purchase share. The inference markup is added to the cost debited from the caller's ORG credit balance and earned via recordCreatorEarnings; the purchase-share applies to the (currently stranded) per-app pool. It is not a one-off invoice. Turning monetization ON is compliance-gated: a newly created app is a review draft, and monetizationEnabled: true is refused with 403 (review_status in the body) until the app reaches approved. Submit it for review first; markup/share values and monetizationEnabled: false are always accepted.
• App charge requests (POST /api/v1/apps/{id}/charges) ask a user to pay an exact USD amount through Stripe or OxaPay. The payer receives app credits; creator earnings flow through the app-credit earnings ledger.
• x402 payment requests (POST /api/v1/x402/requests) ask for direct crypto settlement. Use these when the payer already has crypto or the flow is wallet-native. Current settlement support includes Base, Ethereum, BSC, and Solana; defaults point at https://x402.eliza.app.
• App-credit checkout (POST /api/v1/app-credits/checkout) buys into the per-app pre-purchased credit pool (app_credit_balances). Note: inference billing was migrated to the org balance, so these purchases are currently stranded (issue #8253) — prefer org-credit checkout for spendable balance. Use app charge requests when the agent needs a durable request, metadata, callbacks, and a reusable payment URL.
• Org-credit checkout (POST /api/v1/credits/checkout) tops up the user's organization. It is not creator pricing.
• Cloud tunnel provisioning (POST /api/v1/apis/tunnels/tailscale/auth-key) debits org credits once per successful tunnel auth-key mint. It is on-demand infrastructure usage, not SaaS/subscription billing.
• Redemptions (POST /api/v1/redemptions) request creator payout in elizaOS tokens on base, bsc/bnb, ethereum, or solana. Send asset: "eliza" explicitly; legacy raw omission defaults to USDC and must not be paired with an elizaOS-token quote. Payouts are fixed to the USD quote at request time and then admin reviewed/processed.

For agent-initiated charges, always include callback channel metadata when a conversation should get the payment result:

{ "callback_channel": { "roomId": "room-id", "agentId": "agent-id" } }

On success or failure, the Cloud payment services can write back to that same room so the agent can tell the user whether the payment went through.

When running inside the local @elizaos/plugin-elizacloud route plugin, use /api/cloud/billing/* aliases instead of exposing Cloud credentials to browser or app code. They proxy to the real Cloud API and preserve x402 payment headers:

• /api/cloud/billing/x402/* -> /api/v1/x402/*
• /api/cloud/billing/apps/{appId}/charges/* -> /api/v1/apps/{appId}/charges/*
• /api/cloud/billing/apps/{appId}/earnings/* -> /api/v1/apps/{appId}/earnings/*
• /api/cloud/billing/apps/{appId}/monetization -> /api/v1/apps/{appId}/monetization
• /api/cloud/billing/app-credits/* -> /api/v1/app-credits/*
• /api/cloud/billing/affiliates/* -> /api/v1/affiliates/*
• /api/cloud/billing/redemptions/* -> /api/v1/redemptions/*

Do not hand-calculate payment totals. The creator supplies the requested amount; Cloud returns platform/service fees, total charged amount, headers, URLs, and status fields. Show or store the returned values.

Management surface — what users can ask for

This is the catch-all skill for any user request about apps they already own. Endpoints + intent map:

User says	Endpoint	Method
list my apps	/api/v1/apps	GET
show me my app X / app details	/api/v1/apps/{id}	GET
rename my app / change app config	/api/v1/apps/{id}	PATCH
delete this app	/api/v1/apps/{id}	DELETE
list my containers	/api/v1/containers	GET
change container tier / size	/api/v1/apps/{id} (container fields)	PATCH
what are my earnings	/api/v1/apps/{id}/earnings	GET
set markup percentage	/api/v1/apps/{id}/monetization	PUT (enabling needs review_status: approved, else 403)
charge this user / send a payment request	/api/v1/apps/{id}/charges or /api/v1/x402/requests	POST
check if they paid	/api/v1/apps/{id}/charges/{chargeId} or /api/v1/x402/requests/{id}	GET
create checkout for that charge	/api/v1/apps/{id}/charges/{chargeId}/checkout	POST
create affiliate code	/api/v1/affiliates	POST
link affiliate code	/api/v1/affiliates/link	POST
show payout balance	/api/v1/redemptions/balance	GET
quote payout

Cloud tunnels are multi-tenant by construction: callers must authenticate as an active Cloud user or API key with an organization, provisioning consumes org credits immediately, keys are short-lived/non-reusable/ephemeral, the server forces tag:eliza-tunnel, and the public proxy only forwards generated signed eliza-<org>-<random>-<expiry>-<signature> hostnames into the Headscale tailnet. Signed public hostnames expire with the tunnel provisioning window.

Always confirm before destructive actions (delete app, regenerate key) — show the user what's about to happen, ask for explicit yes.

For domain-specific ops:

• eliza-cloud-buy-domain — register a brand-new domain through cloudflare (paid from cloud credits)
• eliza-cloud-manage-domain — list / edit dns records / detach domains

For the build-and-monetize flow specifically:

• build-monetized-app — ships a new app, then proactively offers a custom domain at the end

Monetization & promotion surfaces (ads + influencers)

Beyond inference markup / purchase share, an app can earn from ads and the agent can promote by hiring influencers — reusing existing credit + earnings rails only (no new payment infra).

Ad inventory / SSP — the app sells ad placements and earns.

User says	Endpoint	Method	Agent action
monetize my app with ads / sell ad space	/api/v1/marketing/inventory	POST	CREATE_AD_SLOT
show my ad slots / ad earnings	/api/v1/marketing/inventory	GET	LIST_AD_SLOTS
pause/edit an ad slot	/api/v1/marketing/inventory/{slotId}	PATCH/DELETE	—
ad slot analytics	/api/v1/marketing/inventory/{slotId}/analytics	GET	—

The public …/inventory/serve?slot= + …/inventory/click endpoints are the miniapp's ad tag (they fill a slot with an eligible campaign, debit the advertiser exactly once, and credit the publisher's redeemable earnings).

Influencer marketplace — publish a profile to earn, or hire influencers to promote.

User says	Endpoint	Method	Agent action
list me as an influencer / offer promo services	/api/v1/marketing/influencers	POST	CREATE_INFLUENCER_PROFILE
find/hire an influencer	/api/v1/marketing/influencers?niche=	GET	LIST_INFLUENCERS
book/hire X to promote for $Y	/api/v1/marketing/influencers/bookings	POST	BOOK_INFLUENCER (money — two-step confirm)
influencer accepts / delivers	…/bookings/{id}/accept | …/deliver	POST	—
advertiser approves (releases escrow) / rejects / cancels (refunds)	…/bookings/{id}/approve | reject | cancel	POST	—

Booking is escrowed: the advertiser's org credits are debited when the offer is funded, released to the influencer on approval, or refunded on reject/cancel (the influencer can decline from offered or accepted). Every money move is idempotent on the booking id and runs before the status finalizes, so a failed payout/refund leaves a retryable state and a retry moves money at most once. BOOK_INFLUENCER never moves money on the first ask; it funds only on an explicit structured confirmation (and sends a per-confirmation idempotency key so a transport retry cannot fund twice).

Together with managed frontend hosting (publish a full app site via DEPLOY_FRONTEND / /api/v1/apps/{id}/frontend), this is the complete agent loop: build → host a full app → deploy backend → attach a domain → monetize (markup / purchase / ads) → promote (influencers) → track earnings → pay out.