# ProxyLLM Blog

> Guides on Codex Hosted, the ChatGPT subscription as an API backend, OpenAI cost math, and the Codex CLI.

## Codex Hosted

- [Codex Hosted Savings: 5 Real Workload Scenarios](https://proxyllm.ai/blog/codex-hosted-savings-examples.md): Five worked examples from a $250 solo dev bill to a $14,000 agent fleet, with full arithmetic. Savings run 40 to 97 percent, and every capacity figure is labeled an estimate.
- [Codex Hosted Setup: From ChatGPT Account to Endpoint in 5 Minutes](https://proxyllm.ai/blog/codex-hosted-setup-guide.md): Sign in to Codex on your machine, connect the account to ProxyLLM with one CLI command, set one base URL, and send your first request. Every step, with commands.
- [Codex Hosted vs Running Codex Yourself: An Honest Comparison](https://proxyllm.ai/blog/codex-hosted-vs-running-codex-yourself.md): DIY Codex on a laptop, VPS, or CI runner is free and fine for personal scripts. The honest ops comparison: uptime, auth refresh, queueing, logs, and the real costs.
- [Codex Hosted: 25 Questions, Answered Directly](https://proxyllm.ai/blog/codex-hosted-faq.md): Twenty-five direct answers about Codex Hosted: what it is, how sign-in works, streaming, models, usage limits, billing, and where OpenAI's terms stand.
- [Fallback Lanes: How Codex Hosted Survives Usage Limits](https://proxyllm.ai/blog/codex-hosted-fallback-lanes-explained.md): Codex Hosted orders your credentials into lanes: ChatGPT accounts first, your API key last. How failover works, what the request log shows, and how to size the stack.
- [How Codex Hosted Billing Works: $129 Flat, No Markup](https://proxyllm.ai/blog/how-codex-hosted-billing-works.md): ProxyLLM charges $129/month flat, no inference markup. OpenAI bills your ChatGPT plan separately. What each invoice covers, what the dashboard tracks, and the free tier's scope.
- [One Account, One Container: How Codex Hosted Isolation Works](https://proxyllm.ai/blog/codex-hosted-architecture-isolated-containers.md): Sign-in on your own machine, no password custody, one isolated container per ChatGPT account, never pooled, AES-256-GCM for stored keys. The Codex Hosted security model, explained.
- [What Is Codex Hosted? ProxyLLM's Main Feature, Explained](https://proxyllm.ai/blog/what-is-codex-hosted.md): Codex Hosted runs OpenAI's official Codex CLI on our servers, signed in with your ChatGPT account, and exposes it as an OpenAI-compatible endpoint billed to your flat plan.
- [What Works with Codex Hosted (and What Doesn't)](https://proxyllm.ai/blog/codex-hosted-for-openai-compatible-apps.md): If a tool accepts an OpenAI base URL, it works with Codex Hosted. The caveats, honestly: complete responses, Codex's model surface, and what stays on your API key.

## ChatGPT subscription as API

- [Can You Turn a ChatGPT Account Into an API Key?](https://proxyllm.ai/blog/turn-chatgpt-account-into-api-key.md): Not into a key: OpenAI issues API keys only on its separate API platform. But a ChatGPT account can become an endpoint, because plans include Codex and Codex runs from code.
- [Can You Use Your ChatGPT Subscription as an API?](https://proxyllm.ai/blog/can-you-use-your-chatgpt-subscription-as-an-api.md): Yes, through OpenAI's Codex CLI. Here is how ChatGPT-plan Codex usage works, what it costs next to the API, and how to expose it as an endpoint.
- [ChatGPT Business and Enterprise: Codex and Programmatic Use](https://proxyllm.ai/blog/chatgpt-business-api-access.md): ChatGPT Business and Enterprise include Codex on every seat, not API credits. How teams run seats programmatically, with admin controls and seat capacity math.
- [ChatGPT Plus vs the API for Coding: 2026 Breakeven Math](https://proxyllm.ai/blog/chatgpt-plus-vs-api-for-coding.md): ChatGPT Plus vs the OpenAI API for coding, at June 2026 GPT-5.x prices: breakeven tables at 1M, 10M, and 100M tokens, and why agent volumes settle the question.
- [ChatMock and DIY Codex Proxies: Setup, Risks, Hosted Alternative](https://proxyllm.ai/blog/chatmock-vs-hosted-codex.md): ChatMock, codex-openai-proxy, and CLIProxyAPI turn a ChatGPT login into a local OpenAI-compatible API. They work. Here is what running one actually involves.
- [Does ChatGPT Plus Include API Access? What You Actually Get](https://proxyllm.ai/blog/chatgpt-plus-api-access-explained.md): ChatGPT Plus includes no API key and no API credits: the API bills separately. But Plus does include Codex, which runs programmatically, an estimated $700 of monthly capacity.
- [Flat-Rate OpenAI API: Does It Exist in 2026?](https://proxyllm.ai/blog/flat-rate-openai-api.md): OpenAI does not sell a flat-rate API. Three workarounds exist: gray-market resellers, DIY Codex proxies, and subscription-backed endpoints. A risk-ranked survey.
- [Is There an API for ChatGPT Pro?](https://proxyllm.ai/blog/is-there-an-api-for-chatgpt-pro.md): No. ChatGPT Pro ships no API key and no credits. It does include Codex with the largest individual plan windows: an estimated $3,500 to $14,000 of monthly API-equivalent capacity.
- [Self-Hosting a Codex Proxy vs ProxyLLM: Total Cost of Ownership](https://proxyllm.ai/blog/diy-codex-proxy-vs-proxyllm.md): A fair TCO comparison: VPS cost, setup hours, auth-refresh breakage, and queueing against a $129 flat fee. With the arithmetic, and the cases where DIY wins.
- [Sign In with ChatGPT: How Subscriptions Became Credentials](https://proxyllm.ai/blog/sign-in-with-chatgpt-explained.md): Sign in with ChatGPT turns a subscription into a credential: where the program started, how the device-code flow works, and which tools can bill work to your plan.
- [The "Unlimited OpenAI API" Myth: What Subscriptions Really Give You](https://proxyllm.ai/blog/unlimited-openai-api-subscription.md): No unlimited OpenAI API exists at any price. What ChatGPT subscriptions actually provide: rolling usage windows, fair-use rules, and how to plan around them.
- [Use Your ChatGPT Subscription in Make.com Scenarios](https://proxyllm.ai/blog/use-chatgpt-subscription-with-make.md): Call a Codex-backed OpenAI endpoint from Make.com with the HTTP module: exact module settings, response mapping, and what an AI-heavy scenario costs both ways.
- [Use Your ChatGPT Subscription in n8n (No API Key Billing)](https://proxyllm.ai/blog/use-chatgpt-subscription-with-n8n.md): Point n8n's OpenAI credential at a Codex-backed base URL and your workflow's AI calls bill to your flat ChatGPT plan. Setup steps, caveats, and the cost math.
- [Use Your ChatGPT Subscription in Zapier Workflows](https://proxyllm.ai/blog/use-chatgpt-subscription-with-zapier.md): Zapier's native OpenAI actions cannot change the API base URL. Webhooks by Zapier can: the exact POST setup that bills your AI steps to a flat ChatGPT plan.
- [Why Agent Workloads Flip the API-vs-Subscription Math](https://proxyllm.ai/blog/chatgpt-subscription-api-for-agents.md): One agent task is 5 to 50 model calls, and context regrows every step. Worked loop math showing where per-token billing breaks and flat plan windows win.
- [Why Your OpenAI API Bill Is Separate from ChatGPT](https://proxyllm.ai/blog/openai-api-billing-separate-from-chatgpt.md): ChatGPT subscriptions and the OpenAI API are two products with two bills. Why your Plus or Pro payment never covers API usage, and where the one documented bridge sits.

## OpenAI costs

- [Estimating Your OpenAI API Costs: A Calculator Walkthrough](https://proxyllm.ai/blog/openai-api-cost-calculator-guide.md): Estimate OpenAI API costs from token counts: 0.75 words per token, the per-request formula, volume with a buffer, then map the total to a ChatGPT plan tier.
- [Fixed-Cost LLM Inference: Options for Predictable AI Bills](https://proxyllm.ai/blog/fixed-cost-llm-inference.md): Three real paths to a flat LLM bill: self-hosted open models on rented GPUs, provider capacity commitments, and subscription-backed Codex. Honest tradeoff table included.
- [GPT-5.5 API Cost: Per-Token Prices and Real Workload Math](https://proxyllm.ai/blog/gpt-5-5-api-cost-per-million-tokens.md): GPT-5.5 costs $5 per million input tokens and $30 per million output as of June 2026. What a PR review, a 100-document run, and an agent session cost, with the math shown.
- [How to Calculate AI Agent Costs (Multi-Step, Tool Calls, Retries)](https://proxyllm.ai/blog/ai-agent-cost-calculation.md): Agents pay for loops, not single completions. The formula: steps x calls x tokens x retry rate, with worked GPT-5 examples and the multipliers public calculators miss.
- [How to Cap OpenAI API Spending Before It Caps You](https://proxyllm.ai/blog/openai-api-spending-limits.md): Set OpenAI's budget alerts and monthly limits, then add hard per-app caps with scoped sub-keys. A runaway GPT-5.5 loop burns about $576 an hour; here is the containment stack.
- [How to Reduce OpenAI API Costs: Every Lever That Works in 2026](https://proxyllm.ai/blog/reduce-openai-api-costs.md): Seven levers that cut OpenAI API bills, with worked numbers for each: model routing, prompt caching, batch jobs, prompt trimming, output caps, monitoring, and a flat-rate lane.
- [LLM Cost per User: The SaaS Math Nobody Shows](https://proxyllm.ai/blog/llm-cost-per-user-saas.md): Cost per user = actions x calls x tokens x price. Worked example: $1.82 at the mean, $18 at p95 against $29 ARPU. Margin bands, and how flat capacity caps tail risk.
- [OpenAI API Pricing Explained (2026): Models, Tokens, Gotchas](https://proxyllm.ai/blog/openai-api-pricing-explained-2026.md): OpenAI API prices per million tokens, June 2026: GPT-5.5 $5/$30, GPT-5 $1.25/$10, Mini $0.25/$2, plus cached input, batch discounts, and the gotchas that inflate bills.
- [OpenAI API vs ChatGPT Subscription: The Real Cost Math](https://proxyllm.ai/blog/openai-api-vs-chatgpt-subscription-cost.md): Per-token API pricing vs flat ChatGPT plans, with worked examples. Where the crossover sits and when a subscription-backed setup wins.
- [OpenAI Costs Suddenly Spiked? Find the Cause in 10 Minutes](https://proxyllm.ai/blog/openai-api-costs-suddenly-high.md): A spike has one of five causes: a loop bug, a model bump, context growth, lost cache hits, or traffic. The 10-minute diagnostic, in order, with math to confirm each.
- [OpenAI Prompt Caching: How It Works and What It Saves](https://proxyllm.ai/blog/openai-prompt-caching-explained.md): OpenAI bills cached input at $0.125 per million tokens on GPT-5 versus $1.25 fresh, as of June 2026. The prefix rules, cache-friendly prompt structure, and hit-rate math.
- [Reasoning Tokens: The Hidden Multiplier in Your OpenAI Bill](https://proxyllm.ai/blog/openai-reasoning-tokens-cost.md): Reasoning tokens bill as output tokens at full price. A 150-token answer can bill 1,500 tokens, turning a $525 monthly estimate into $2,550. How to see and control them.
- [The Cheapest OpenAI Model That Still Does the Job](https://proxyllm.ai/blog/cheapest-openai-model-for-production.md): GPT-5 Nano at $0.05/$0.40 per million tokens handles classification and extraction; Mini covers most production text. The decision table, with a worked 94x cost spread.
- [The OpenAI Batch API: When 50% Off Is Worth the Wait](https://proxyllm.ai/blog/openai-batch-api-discount-explained.md): The Batch API takes 50 percent off OpenAI token prices for jobs returned within 24 hours. How it works, which workloads qualify, and when a flat subscription lane beats it.
- [What a 24/7 AI Agent Actually Costs: API vs Subscription](https://proxyllm.ai/blog/what-a-24-7-ai-agent-costs.md): A 30-day worked scenario at three volumes: $239 to $7,155 a month on metered GPT-5 pricing, against flat ChatGPT plan setups at $149 to $329. Estimates, math shown.
- [Why Is My OpenAI Bill So High? A Diagnostic Checklist](https://proxyllm.ai/blog/why-is-my-openai-bill-so-high.md): Eight causes explain almost every oversized OpenAI bill: reasoning tokens, context bloat, agent loops, retries, wrong model, no caching, uncapped output, forgotten keys.

## Codex CLI

- [AGENTS.md: Steering Codex with a Repo Contract](https://proxyllm.ai/blog/codex-agents-md-guide.md): AGENTS.md is the instruction file Codex reads before working in your repo. What belongs in it, what to leave out, and a complete example for an exec-driven repository.
- [Can GitHub Actions Use Your ChatGPT Plan for Codex?](https://proxyllm.ai/blog/codex-github-actions-chatgpt-plan.md): Mechanically possible, but the official action expects an API key, and a personal plan session on shared runners blurs OpenAI's account rules. Three clean patterns, compared.
- [Codex Auth: API Key vs ChatGPT Sign-In, Compared](https://proxyllm.ai/blog/codex-api-key-vs-chatgpt-sign-in.md): ChatGPT sign-in bills Codex to your flat plan with usage windows. An API key bills per token with per-minute limits. How to choose, and when to run both as lanes.
- [Codex CLI in Docker: A Working Container Setup](https://proxyllm.ai/blog/codex-cli-docker-guide.md): A Dockerfile for OpenAI's Codex CLI, a named volume for ~/.codex auth, device login inside the container, and why the sandbox flag changes when Docker is the boundary.
- [codex exec Cookbook: 12 Non-Interactive Recipes](https://proxyllm.ai/blog/codex-exec-cookbook.md): Twelve copy-paste codex exec recipes: PR review, TODO extraction, doc and changelog generation, cron digests, webhook handlers, jq pipelines, and resumed multi-stage chains.
- [codex exec: The Complete Guide to Non-Interactive Codex](https://proxyllm.ai/blog/codex-exec-non-interactive-mode-guide.md): How OpenAI's codex exec works: syntax, auth on headless machines, CI usage, resume, and when a hosted container beats running it yourself.
- [Codex in CI/CD: Pipelines, Gates, and Non-Interactive Runs](https://proxyllm.ai/blog/codex-cli-ci-cd-guide.md): How to run OpenAI's Codex CLI in CI/CD: codex exec for pre-merge checks, sandbox and approval presets, output piping, and the right auth for shared vs personal runners.
- [Codex Limits: ChatGPT Plus vs Pro vs Business](https://proxyllm.ai/blog/codex-limits-plus-vs-pro.md): Every ChatGPT plan includes Codex; the window sizes differ. How Plus, Pro, and Business capacity compares, which tier fits which workload, and the cost math as estimates.
- [codex login Without a Browser: Device Auth Step by Step](https://proxyllm.ai/blog/codex-login-without-browser.md): Run codex login --device-auth on the headless box, approve the code at chatgpt.com from any device, and the CLI saves its tokens to ~/.codex/auth.json. Full walkthrough.
- [Codex On-Demand Credits: Pricing, Draining, and When to Skip Them](https://proxyllm.ai/blog/codex-credits-explained.md): Codex credits let Plus and Pro users keep working after a limit. How they bill, why users report fast drain on agent work, and when a second account is the cheaper valve.
- [Codex Rate Limit Errors: Causes and Fixes](https://proxyllm.ai/blog/codex-rate-limit-errors-fix.md): A Codex rate limit error means one of four things: the 5-hour window, the weekly cap, drained credits, or an API 429. How to tell which one, and the fix for each.
- [Codex Usage Limits, Explained (Updated June 2026)](https://proxyllm.ai/blog/codex-usage-limits-explained.md): How Codex limits work: rolling 5-hour windows, weekly caps on some plans, one shared usage pool, and on-demand credits. The full system, with links to OpenAI's live numbers.
- [Codex with Multiple ChatGPT Accounts: Strategies That Work](https://proxyllm.ai/blog/codex-multiple-accounts.md): The Codex CLI signs in one ChatGPT account at a time. The workarounds that exist: separate CODEX_HOME directories, re-login juggling, and gateway fallback lanes.
- [Codex's 5-Hour Window: What Counts and What Resets](https://proxyllm.ai/blog/codex-5-hour-limit-explained.md): Codex meters usage over a rolling window of roughly five hours. What draws it down, including reasoning time, how resets behave, and how to pace heavy sessions.
- [How to Check Your Codex Usage and Remaining Limits](https://proxyllm.ai/blog/check-codex-usage.md): Check Codex usage with /status in the CLI, ChatGPT's usage surfaces, or per-request gateway logs. Where to see your 5-hour and weekly limits and when they reset.
- [OpenAI Codex CLI: Getting Started in 10 Minutes](https://proxyllm.ai/blog/codex-cli-getting-started.md): Install OpenAI's Codex CLI with npm or Homebrew, sign in with your ChatGPT account, run your first task, and set sane defaults. A practical setup guide.
- [Run Codex CLI Headless: Servers, SSH, and No-Browser Auth](https://proxyllm.ai/blog/codex-cli-headless-server-setup.md): Run Codex CLI on a headless server: install, device-code auth with no browser, codex exec for scripts, auth.json handling, and systemd patterns that survive reboots.
- [Run Codex on a Schedule: Cron, Webhooks, and Always-On Jobs](https://proxyllm.ai/blog/run-codex-on-a-schedule.md): Schedule codex exec with cron or systemd timers, wire failure alerts with exit codes and webhooks, and pick the job's home: laptop, VPS, or a hosted endpoint.
- [Running Codex CLI on a VPS: The Complete Setup](https://proxyllm.ai/blog/codex-cli-on-vps.md): Rent a $5 box, install Codex CLI, sign in with device auth, keep it alive with tmux or systemd, and guard auth.json. The full VPS walkthrough, plus the honest hosted bridge.
- [The Codex GitHub Action: PR Reviews and Quality Gates](https://proxyllm.ai/blog/codex-github-action-guide.md): Set up OpenAI's official Codex GitHub Action with an OPENAI_API_KEY repo secret, post PR reviews as comments, and gate merges on a Codex verdict. Working YAML included.
- [The Codex Weekly Limit: How It Works and How to Live With It](https://proxyllm.ai/blog/codex-weekly-limit-explained.md): Some ChatGPT plans cap Codex over a 7-day cycle on top of the 5-hour window. How the weekly limit resets, why agent work burns it fast, and how to keep shipping.

## Comparisons

- [Are Cheap OpenAI API Resellers Legit? The Gray Market, Examined](https://proxyllm.ai/blog/openai-api-resellers-legit.md): Most cheap OpenAI API resellers run on bulk accounts and resold keys OpenAI's terms prohibit. How the gray market sources capacity, how it ends, and the clean alternative.
- [Cheap OpenAI API Access: Every Path, Ranked by Risk](https://proxyllm.ai/blog/cheap-openai-api-access.md): Five ways to pay less for OpenAI: official discounts, subscription-backed capacity, DIY proxies, gray-market resellers, and shared keys, ranked by what can go wrong.
- [Codex vs Claude Code: The Cost Structure Comparison](https://proxyllm.ai/blog/codex-vs-claude-code-cost.md): Codex bills through flat ChatGPT plan windows; Claude Code meters programmatic use through Agent SDK credits or API keys. The structures and policies, compared.
- [Helicone Alternatives After the Mintlify Acquisition](https://proxyllm.ai/blog/helicone-alternatives.md): Helicone joined Mintlify in 2026, and teams are rechecking their LLM observability options: Langfuse, Portkey, gateway built-in logs, and lightweight request logging.
- [LiteLLM vs OpenRouter: Self-Hosted Router or Hosted Marketplace?](https://proxyllm.ai/blog/litellm-vs-openrouter.md): LiteLLM is an open-source router you run yourself, with no markup. OpenRouter is a hosted marketplace charging about 5.5% on credits. How to choose on fees, ops, and models.
- [NanoGPT Review: What $8/Month Flat Actually Buys](https://proxyllm.ai/blog/nano-gpt-review.md): An honest NanoGPT review: the $8/month flat tier is real value for individuals on open-weight models, with fair-use limits on pooled capacity. Who it fits and who it doesn't.
- [NanoGPT vs OpenRouter: Flat Subscription Meets Pay-Per-Token](https://proxyllm.ai/blog/nano-gpt-vs-openrouter.md): NanoGPT sells a flat $8/month tier on capacity pooled across its own accounts; OpenRouter meters 400+ models per token plus a ~5.5% credit fee. How to pick between them.
- [OpenRouter Alternatives in 2026 (Including the Flat-Rate Option)](https://proxyllm.ai/blog/openrouter-alternatives.md): Five OpenRouter alternatives compared on the axis most lists skip: cost model. LiteLLM, Portkey, Requesty, direct provider keys, and ProxyLLM's flat-rate lane.
- [OpenRouter vs Direct OpenAI: When the Middleman Earns His Fee](https://proxyllm.ai/blog/openrouter-vs-direct-openai.md): OpenRouter adds roughly 5.5% on credits and a BYOK fee on top of OpenAI's list prices. The fee math at three volumes, what the marketplace buys, and when direct keys win.
- [Per-Token vs Flat-Rate LLM Pricing: A Decision Framework](https://proxyllm.ai/blog/per-token-vs-flat-rate-llm-pricing.md): Per-token billing prices variance; flat plans price a ceiling. A decision table by utilization, burst shape, and risk tolerance for picking an LLM pricing model.
- [Portkey Alternatives: Gateways, Routers, and Flat-Rate Lanes](https://proxyllm.ai/blog/portkey-alternatives.md): Portkey bundles a gateway, observability, and guardrails. The real alternatives sorted by job: LiteLLM, OpenRouter, Langfuse, Helicone, and a flat-rate lane for OpenAI spend.
- [ProxyLLM Alternatives: DIY Proxies, Gateways, and Going Direct](https://proxyllm.ai/blog/proxyllm-alternatives.md): The honest list: ChatMock or CLIProxyAPI if you self-host, OpenRouter for model breadth, LiteLLM for routing, the direct API for low spend. And who should not buy from us.
- [ProxyLLM vs LiteLLM: Hosted Flat-Rate vs Self-Hosted Router](https://proxyllm.ai/blog/proxyllm-vs-litellm.md): LiteLLM is a free self-hosted router across many providers. ProxyLLM is a hosted flat-rate lane for OpenAI volume. Different layers, and they compose well.
- [ProxyLLM vs OpenRouter: Different Problems, Different Tools](https://proxyllm.ai/blog/proxyllm-vs-openrouter.md): OpenRouter sells model breadth per token; ProxyLLM sells flat OpenAI volume on your own ChatGPT plan. What each is for, when to run both, and who should buy neither.
- [The Best LLM Gateways in 2026, by Workload](https://proxyllm.ai/blog/best-llm-gateway-2026.md): No single best LLM gateway exists; workloads pick winners. OpenRouter for multi-model apps, LiteLLM for platform teams, Portkey for observability, ProxyLLM for flat OpenAI volume.

## Integrations

- [Controlling AI Costs in GitHub Actions](https://proxyllm.ai/blog/github-actions-ai-costs.md): CI runs AI on every push, with retries and matrix jobs multiplying calls. Capped per-repo keys, request logs per key, and when CI belongs on a key vs a plan.
- [Cursor and Your ChatGPT Subscription: What Works in 2026](https://proxyllm.ai/blog/cursor-chatgpt-subscription.md): Cursor's OpenAI key override can point chat at a subscription-backed endpoint. What survives the override, what stays on Cursor's stack, and when it pays off.
- [Dify Apps Without Metered OpenAI Billing](https://proxyllm.ai/blog/dify-openai-cost-control.md): Add ProxyLLM as an OpenAI-compatible provider in Dify and app traffic bills to a flat ChatGPT plan. Per-app sub-keys with hard caps make each app's spend visible.
- [Flowise Agent Flows on Flat OpenAI Capacity](https://proxyllm.ai/blog/flowise-openai-cost-control.md): Point Flowise's ChatOpenAI credential at a flat OpenAI lane: the config, worked cost math for an agent flow that loops and retries, and caps per deployment.
- [LangChain on a Flat-Rate OpenAI Lane](https://proxyllm.ai/blog/langchain-flat-rate-openai.md): ChatOpenAI accepts a base_url, so LangChain chains and agents can bill to a flat ChatGPT plan through Codex Hosted. Python and JS setup, plus worked agent-loop math.
- [LlamaIndex Ingestion Without the Token Meter](https://proxyllm.ai/blog/llamaindex-openai-costs.md): LlamaIndex ingestion is a bulk LLM workload: three extractors over 10,000 chunks is 30,000 model calls. The api_base setup and the worked math, metered vs flat.
- [MCP Tool Loops and the Cost of Agency](https://proxyllm.ai/blog/mcp-server-llm-costs.md): MCP hosts re-send every tool schema with every model call, and each tool result triggers another call. The worked math of tool-loop spend and what a flat lane changes.
- [Node.js OpenAI SDK Behind a Flat-Rate Endpoint](https://proxyllm.ai/blog/nodejs-openai-base-url-proxy.md): Point the OpenAI Node.js SDK at ProxyLLM with baseURL or env vars. Services and serverless functions bill OpenAI calls to a flat ChatGPT subscription.
- [opencode on Your ChatGPT Subscription via ProxyLLM](https://proxyllm.ai/blog/opencode-chatgpt-subscription.md): Add ProxyLLM to opencode.json with @ai-sdk/openai-compatible and coding sessions bill to your ChatGPT plan. The config block, session cost math, and limit behavior.
- [Python OpenAI SDK: One base_url Change for Flat-Rate Calls](https://proxyllm.ai/blog/python-openai-base-url-proxy.md): How to point the OpenAI Python SDK at ProxyLLM with base_url. Batch scripts and notebooks bill to a flat ChatGPT subscription instead of the per-token meter.
- [The OpenAI-Compatible API, by Hand: curl Examples](https://proxyllm.ai/blog/curl-openai-compatible-api.md): Raw curl requests against an OpenAI-compatible endpoint: headers, request body, response shape, and error envelope, with flat-rate billing behind the URL.
- [Vercel AI SDK with ProxyLLM: Provider Setup and Caveats](https://proxyllm.ai/blog/vercel-ai-sdk-proxyllm-setup.md): Configure createOpenAICompatible with ProxyLLM's base URL and OpenAI calls bill to a flat ChatGPT plan. Setup code, the streaming caveat, and per-environment keys.

## AI agency

- [AI Agency Pricing Models: 5 Structures with Real Numbers](https://proxyllm.ai/blog/ai-agency-pricing-models.md): Retainer, pass-through, capped, value-based, and hybrid pricing for AI agency work, each with worked numbers and its failure mode, plus what flat COGS changes.
- [AI Agency Unit Economics: Where Margin Actually Leaks](https://proxyllm.ai/blog/ai-agency-unit-economics.md): Revenue per client is flat; cost per client is metered. The agentic margin concept, a quantified leak inventory (retries, scope creep, model bumps), and the math to plug it.
- [Forecasting OpenAI Costs for Client Projects](https://proxyllm.ai/blog/forecast-openai-costs-client-projects.md): A working estimation method for agencies: prototype a thin slice, sample real token counts, multiply by steps and retries, quote a range, then track it weekly.
- [Should You Pass API Costs Through to Clients?](https://proxyllm.ai/blog/pass-through-api-costs-to-clients.md): Pass through when usage is client-driven, absorb when it is small and stable, cap when clients need certainty. Failure modes, sample contract language, and the flat-capacity case.
- [The AI Agency Margin Problem: When OpenAI Is Your Biggest Vendor](https://proxyllm.ai/blog/ai-agency-margins-openai-api.md): API costs scale with every client you add. How agencies price AI work, where margins leak, and what flat subscription-backed capacity changes.
- [The Real Cost of an AI Agency Tech Stack (2026)](https://proxyllm.ai/blog/ai-agency-tech-stack-costs.md): Fixed tools run $50-420 a month. The model line is the only one that scales with clients, and at 25 clients it is about 89% of the stack. The itemized table and how to flatten it.
- [White-Label AI Infrastructure for Agencies](https://proxyllm.ai/blog/white-label-ai-agency-infrastructure.md): White-label AI for agencies usually means chatbot builders. The margin layer sits below: a sub-key per client, budget caps, and cost reports that carry your brand.

## Policies & limits

- [Anthropic's Agent SDK Credits: The June 2026 Rules, Explained](https://proxyllm.ai/blog/anthropic-agent-sdk-credits-explained.md): Anthropic now includes metered Agent SDK credits with Claude plans: $20 on Pro, $100 on Max 5x, $200 on Max 20x. What they permit, what stays prohibited, and the math.
- [Is Codex Hosted Against OpenAI's Terms? An Honest Reading](https://proxyllm.ai/blog/is-codex-hosted-against-openai-terms.md): codex exec is documented functionality and Codex is included in ChatGPT plans. What the terms say, what they don't, and why OpenAI keeps the final call.
- [Sharing an OpenAI Account: What the Terms Actually Say](https://proxyllm.ai/blog/is-sharing-openai-account-against-tos.md): Yes, it is prohibited: OpenAI's terms say you may not share credentials or make your account available to anyone else. What that covers, and why owning two accounts differs.
- [Third-Party Tools and Your OpenAI Account: A Risk Hygiene Guide](https://proxyllm.ai/blog/openai-account-ban-risks-third-party-tools.md): OpenAI restricts accounts for shared credentials, scraping-style extraction, and resold access, not for tool choice. The six-question checklist to vet any tool, including us.
- [What Happens When You Hit Your Codex Usage Limit?](https://proxyllm.ai/blog/what-happens-when-you-hit-codex-usage-limit.md): ChatGPT plan limits reset on rolling windows. How limits behave, and how multi-account and API-key fallback keep production requests flowing.
- [Why We Don't Support Claude Code (And Won't)](https://proxyllm.ai/blog/why-no-claude-code-support.md): Anthropic prohibits third-party services from routing requests through Claude subscription credentials. The policy, the January 2026 crackdown, and what it means.