Weekly API Usage & Cost Brief — 2026-07-24

Reporting window: Jul 17, 2026 12:09 PM ET → Jul 24, 2026 12:09 PM ET

Executive Summary

• Measured weekly cost from APIs that returned dollar data: $0.55
• Projected 4.33-week monthly run rate from measured services: $2.39
• Week-over-week measured change: Baseline week
• Automated telemetry: 4 providers; dashboard/manual coverage: 17 providers; errors/gaps: 1
• Important: the measured total is a floor, not the complete technology bill, whenever subscription/OAuth providers or dashboard-only services are present.

Provider Breakdown

| Provider | Category | Status | Weekly cost | Usage / balance |

|---|---|---:|---:|---|

| OpenAI API | LLM/image | manual | Unavailable | HTTP 403 — organization cost API generally requires an Admin API key. ChatGPT/Codex subscription usage is tracked separately. |

| OpenClaw model sessions | Agent runtime | live_proxy | Unavailable | 2 model line(s) |

| KIE.ai | AI media | live | Unavailable | 1304.79 credits remaining |

| Fal.ai | AI media | error | Unavailable | HTTP 403 |

| ElevenLabs | Audio/voice | live | Unavailable | 685 / 90,000 characters/credits |

| Retell AI | Voice agents | live | $0.55 | 2 calls / 3.14 min |

| Anthropic Claude CLI | Agent runtime | manual | Unavailable | OAuth/subscription route; check Claude plan limits. No reliable dollar-usage API connected. |

| Google Gemini | LLM/multimodal | manual | Unavailable | API key connected; usage/cost depends on AI Studio or Google Cloud billing project. |

| WaveSpeedAI | AI media | manual | Unavailable | API key connected; no verified account usage endpoint in the local adapter. |

| HeyGen | Avatar video | manual | Unavailable | Connected app can expose current credits, but recurring CLI billing telemetry is not yet available. |

| OpenArt | AI media | manual | Unavailable | Usage is dashboard/credit based; no local billing API connection. |

| Shotstack | Video rendering | manual | Unavailable | Usage is credit based; no active billing telemetry key was found in the standard provider env. |

| Apify | Research/scraping | manual | Unavailable | Usage depends on connected account and actor runs; no local billing collector configured. |

| Firecrawl | Research/scraping | manual | Unavailable | Search connector is available; billing/credit endpoint is not connected to this report. |

| Blotato | Social publishing | manual | Unavailable | Publishing API connected; plan/usage billing is dashboard based. |

| Twilio | Telephony | manual | Unavailable | Credentials exist for Retell routing; account balance/usage collector is not enabled. |

| Neon | Database | manual | Unavailable | Connected database; review compute/storage against plan in Neon dashboard. |

| Airtable | Database/ops | manual | Unavailable | API connected; track rate/billing limits in Airtable account dashboard. |

| GoHighLevel | CRM/marketing | manual | Unavailable | API connected; subscription and rebilling usage are dashboard based. |

| Cloudinary | Media storage | manual | Unavailable | Connector available; account quota telemetry is not included yet. |

| Shopify | Commerce | manual | Unavailable | Admin API connected; subscription/app billing is dashboard based. |

| Printify / Printful | Fulfillment | manual | Unavailable | APIs connected; production charges are order based, not token/credit usage. |

Detailed Usage

OpenClaw model sessions

• anthropic/claude-haiku-4-5 — 1 session(s); 10 input; 3 output; 18,236 cache-read tokens
• openai/gpt-5.6-sol — 9 session(s); 1,068 input; 87 output; 112,384 cache-read tokens

Risks, Alerts & Actions

• Fal.ai: HTTP 403
• Review any provider whose remaining credits fall below one normal production batch.
• Investigate weekly cost increases above 25% unless tied to an approved campaign or production run.
• Keep failed generations and retries visible; they can burn credits without producing usable assets.
• Separate subscription/OAuth model usage from metered API spend to avoid treating token volume as an invoice.

Coverage Notes

• Dollar totals include only providers that returned dated cost records.
• Token counts from OpenClaw are operational snapshots, useful for model-load trends but not exact billing records.
• Dashboard-only services remain listed so they can be upgraded to automated telemetry later.
• No generation, publishing, account modification, purchase, or spend action was performed by this report.