CodingPlanX logo
Api Tools · Llm Providers

CodingPlanX

CodingPlanX is an API gateway giving one key access to AI models from many vendors. It advertises 600+ models while its catalogue lists 421, speaks the OpenAI protocol, and targets developers, startups and enterprise teams cutting inference costs.

Active Pay As You Go API available 18+ Verified by Guidaio
Overview

What is CodingPlanX?

CodingPlanX is an API gateway sold as a single door onto the wider LLM market: one key, one balance, and calls forwarded to models from many different vendors. The site's own title advertises one AI gateway for 600+ models, the documentation hub states AI Models 100+, and the public catalogue announced 421 models across multiple vendors and groups when it was read on 10 August 2026 - three different figures, all published by the site itself, which is worth knowing before quoting any of them.

The catalogue is filterable by vendor (OpenAI, DeepSeek, Anthropic, Moonshot, Google, Vidu and sixteen others), by tag, by group and by billing type, and each entry shows an input and an output price. The terms list the families reached: Google Gemini, OpenAI GPT-4, GPT-4o, GPT-5 and o4 Mini, Anthropic Claude, xAI Grok, DeepSeek, Moonshot Kimi, Qwen3, Z.AI GLM and Ant Group Bailing. Text is not the whole story. The documentation also covers chat and responses, drawing models, video models, text-to-music through Suno, reranking, and aggregation platforms such as Kling, Replicate, Hailuo Minimax, VIDU and Fal-ai. Token-billed entries are quoted per million tokens, image and video entries per call.

Plugging in is deliberately unremarkable. The service follows the OpenAI standard, so an existing client only needs its base URL and key changed; Anthropic and Vertex protocols are available too, and a Web UI is mentioned. The homepage shows a POST to the chat completions endpoint with a bearer token and an entirely familiar request body.

The pitch is price: up to 90% cost savings, inference at a fifth of the price, and Save-91% banners in the catalogue. The comparison table on the homepage sets the product against categories rather than named competitors, namely traditional proxies and official APIs.

What is not said matters as much. The operator, DeepNail Limited, publishes no address; the domain was registered on 7 March 2026 and first archived on 8 May 2026; every page canonicalises to codingplanx.com rather than the .ai actually served; the blog is empty; and the nine homepage testimonials carry template names.

What it does

  • Call AI models from many different vendors through a single API key and a single balance.
  • Keep existing code untouched: only the base URL and the key change, the endpoint following the OpenAI standard, with Anthropic and Vertex protocols also offered.
  • Power coding agents: Claude Code, Cursor, Cline, Codex, OpenCode, OpenClaw, Gemini CLI and Qwen Code.
  • Generate text, images, video, audio and music, and run reranking, through the same gateway.
  • Let an Auto mode handle routing, or pin a specific model from the catalogue.
  • Follow spending through per-model published pricing and the detailed usage logs the homepage advertises.
  • Buy credits on demand, subscribe monthly or yearly, and optionally switch on automatic card deduction.
Audience

When to use CodingPlanX / When not to

A quick filter to help you decide if CodingPlanX is the right fit.

When to use CodingPlanX

  • Developers already calling an OpenAI-compatible endpoint, who only want to swap the base URL and the key to reach other vendors.
  • Users of coding agents such as Claude Code, Cursor, Cline, Codex, OpenCode, OpenClaw, Gemini CLI or Qwen Code, who would rather have one billing account behind all of them.
  • Freelance developers who cannot justify paying for several premium model subscriptions side by side.
  • Small teams that want to compare a handful of models before committing, without a large budget.
  • Enterprise teams ready to ask for the announced dedicated support, SLA, volume discounts and custom API integration.

When not to use CodingPlanX

  • Organisations handling regulated or personal data: no DPA, no security certification, no subprocessor list, data stored in Singapore with international transfers expressly foreseen.
  • Teams that must guarantee their prompts are never used for training: the privacy policy announces training on anonymised user content and documents no way to opt out.
  • Non-technical users looking for a ready-made interface: everything runs through an API key, and the Web UI the site mentions is neither documented nor shown.
  • Anyone wanting to work from a phone or a browser add-on: no mobile application and no browser extension exist.
  • Buyers whose procurement needs an identifiable counterparty: DeepNail Limited publishes no postal address, no named officer and no beneficial owner, and elects Singapore law with exclusive jurisdiction in Singapore.
Get started

How to use CodingPlanX

A typical end-to-end flow, from setup to results.

  1. Create an account: the Start for free button leads to the registration page, and registration is required by the terms.
  2. Generate an API key from the user area, reachable through the Get API key button on the homepage.
  3. Pick a model in the catalogue using the vendor, tag, group and billing-type filters, or leave the Auto mode to route requests.
  4. Replace the base URL of your tool with the CodingPlanX API endpoint and paste the key in place of the previous one.
  5. Call the chat completions endpoint in OpenAI format, following the curl and PHP samples on the homepage, or use the Anthropic or Vertex protocols.
  6. For a coding agent, follow its dedicated documentation page: Claude Code, Cursor, Codex, OpenCode, OpenClaw or Gemini CLI each have one.
  7. Buy credits on the pricing page, from the USD 10 pack up to the USD 5,000 pack, or take a monthly or yearly subscription.
  8. Optionally switch on automatic card deduction, described in the withholding agreement page, and turn it off again from Payment Settings in the console.
  9. Track consumption and subscriptions from the user area.
  10. Write to contact@codingplanx.ai for technical support, an enterprise request or a refund.
Quick read

Pros & Cons

Pros

  • One key and one balance for models coming from a long list of different vendors.
  • Prices published openly, model by model, for input and for output tokens.
  • Savings advertised between 79% and 91% depending on the model, shown as banners in the catalogue.
  • No code change required thanks to OpenAI standard compatibility, with Anthropic and Vertex as alternatives.
  • A dedicated integration page for each supported coding agent, six in total.
  • Entry at USD 10 with no commitment, and credits usable across the whole catalogue.
  • Refunds possible within 24 hours, minus a 5% platform fee, and card details kept only as a token by the payment gateway.

Cons

  • The site contradicts itself on its central number: 600+ models on the homepage, 100+ on the documentation page, and 421 actually listed in the catalogue on 10 August 2026.
  • No mention of the GDPR, no DPA, no security certification and no subprocessor list, only categories of third parties.
  • Training on anonymised user content is announced, with no documented way to exclude your data from it.
  • The card withholding agreement, a contractual document, still contains unreplaced template markers such as [Product/Service Name], [Your Customer Service Email], [X] hours and [Your Company's Country/Region, e.g., the People's Republic of China].
  • Other finishing defects run through the site: the English contact page ends with a Chinese translation note left in by mistake, the nine homepage testimonials use template names such as Leslie Alexander, Tom Cook and Floyd Miles, every page canonicalises to codingplanx.com instead of the .ai served, the blog is online but empty, and there is neither an about page nor a LinkedIn account.
  • The publisher is barely documented: no postal address, no named officer, no identifiable beneficial owner, with company and domain both surfacing in 2026.
  • A very short refund window of 24 hours with a 5% fee retained and premium pay-as-you-go credits forfeited, and no public SLA outside the enterprise offer even though Uptime 99.99% is displayed.
Pricing

Pricing & Plans

No permanent free plan is documented and no free trial is announced; free credits can only be earned through the referral programme. The cheapest paid entry point is the Lite credit pack, 10 credits for USD 10, with no commitment, the range extending to USD 5,000 with credit bonuses of 5% to 14% on the larger packs. A Subscription tab exists on the pricing page and the terms name a CodingPlanX Builder Plan, but no subscription amount is published, so a cheaper monthly entry point may exist. Consumption is billed either per million tokens or per call depending on the model. Amounts are in US dollars and taxes are not included.

Plan 1
Lite
  • 10 credits - USD 10
Plan 3
Basic
  • 50 credits - USD 50
Plan 4
Plus
  • 105 credits - USD 100 (5% bonus)
Plan 5
Pro
  • 212 credits - USD 200 (6% bonus)
Plan 6
Max
  • 550 credits - USD 500 (10% bonus)
Plan 7
Ultra
  • 1100 credits - USD 1
  • 000 (10% bonus)
Plan 8
Ultimate
  • 5750 credits - USD 5
  • 000 (14% bonus)
Every pack advertises the same inclusions
  • 600+ top-tier models
  • code
  • image
  • video and chat
  • API requests
  • and unlimited concurrency.
Monthly or yearly subscription
  • announced on the pricing page but no amount is published there
  • the terms name a CodingPlanX Builder Plan without a price.
Enterprise, on request
  • dedicated support and an SLA
  • flexible pricing with volume discounts
  • custom integration and technical consulting.
Special offers — Credit bonuses on the larger packs: 5% on Plus, 6% on Pro, 10% on Max, 10% on Ultra and 14% on Ultimate. · Referral programme: USD 0.50 in credits for sharing on social networks, USD 0.50 for each friend who registers, and 30% commission when a referred friend subscribes. · Enterprise offer on request, with flexible pricing and volume discounts. · No student, teacher, non-profit or jobseeker offer is announced anywhere on the site.
Prices and plans listed above may evolve. Always check the official pricing page before subscribing.
Trust & Privacy

Data, GDPR & hosting

A consolidated view of how CodingPlanX handles your data.

GDPR overview

There is no mention of the GDPR anywhere on this site. Neither GDPR nor CCPA nor Article 27 appears in any of the pages collected; no EU representative is designated, no data protection officer is named, and the privacy policy refers only to applicable laws and regulations without naming any. What stands in its place: section VI announces rights of access, update and correction, exercised by email to contact@codingplanx.ai after identity verification, and states that users may have the right to lodge a complaint with their supervisory authority. Section II foresees international transfers and storage outside the user's country of residence, requiring a comparable level of protection from the recipient; data is otherwise held in Singapore. Section IX disclaims any intention to collect data from minors. The terms elect Singapore law and the exclusive jurisdiction of Singapore courts. No DPA is published.

Who owns the data?

DeepNail Limited, the operator, holds every intellectual property right over the service itself: software, technology, models, code, interfaces and documentation (terms 6.1). The user keeps ownership and the relevant rights over the content they submit (terms 6.2), and the privacy policy confirms that copyright in the input stays with them; input and output together are treated as User Content. The publisher nevertheless reserves the right to store, transfer or train on that content once anonymised, to improve its model. Sharing is described by category rather than by name: affiliates and partners, service providers and processors, authorised agents, professional advisers, public authorities, and any counterparty to a merger, acquisition, reorganisation or financing. No subprocessor is ever named.

Reuse rights

Nothing in the terms assigns the outputs to the publisher, so what a user generates is theirs to reuse without asking permission from CodingPlanX. Three reservations sit on top of that freedom. First, the upstream providers bind the user directly: terms 2.3 requires compliance with the terms of OpenAI, Anthropic, Google, xAI, DeepSeek, Moonshot, Qwen, Z.AI and Ant Group, including their usage restrictions, intellectual property clauses and data protection policies. Second, the user warrants that their input is original or lawfully licensed (terms 6.3). Third, outputs may simply be wrong: terms 7.1 places the duty to evaluate them, and the resulting risk, on the user. Two further limits concern the service rather than the content: it cannot be resold, rented, sublicensed or transferred (terms 4.2.3), and the company's trademarks and domain names may not be used (terms 6.4). Subject to the user's authorisation, the publisher may use submitted content to provide, maintain, develop and improve the service (terms 6.2).

Data retention & training

Retention summary
No duration is ever put in figures. The privacy policy says only that personal information is kept for the period necessary for the purposes set out in the policy and within the timeframe required by applicable law. When use of the service ends, collection stops unless a legal obligation applies; where a statutory retention period is still running, the data is kept but no longer processed, apart from storage and security measures. If the business ceases, collection stops, users are informed, and the data held is deleted or anonymised. Anonymised user content is treated separately and may be stored, transferred or used to train models. Account deletion can be requested, subject to approval, and access, update or correction go through the published contact address after identity verification. No retention period is given for API call logs.
Trains on customer data
Yes
GDPR contact

Hosting summary

Jurisdiction: the terms elect Singapore law and the exclusive jurisdiction of the Singapore courts. Storage: the privacy policy states that, generally, all data related to the account and to the product services, personal information and user content alike, is stored in Singapore. Transfers: the same section expressly provides for storage or transfer outside the user's country of residence, requiring a comparable level of protection from the recipient. No region, data centre or hosting provider is named beyond the country itself, and third parties appear only as categories, never as a named subprocessor list. Announced measures include strictly restricted data centre access, private network channels and proxies, staff dedicated to data protection, and contingency plans with regular drills. Card details are received and stored encrypted by third-party gateways certified PCI-DSS, the publisher keeping only a token and the last four digits. On the network side, the domain resolves to a Cloudflare anycast address, which is a CDN point of presence and not a storage location. No SOC 2 or ISO 27001 certification is mentioned anywhere on the site.

Hosting countries
🇸🇬 Singapore
Watch-outs

Things to keep in mind

Risks and trade-offs to weigh before adopting CodingPlanX.

  • A single API key opens the door to unlimited spending, and the terms put any leak on the user: malicious calls made with a leaked key are explicitly not refunded.
  • Card deduction is described as automatic and password-free, triggered from 5 CNY or USD outstanding, which is easy to forget and hard to notice.
  • User content may be used to train models once anonymised, and no exclusion mechanism is documented, so the decision is made for you by default.
  • Outputs can simply be wrong: the terms warn that the service may generate inaccurate code or information, while comfortable pricing encourages unmeasured model consumption and a slow handover of your own judgement to the tool.
  • The gateway adds one more intermediary to the chain, since prompts pass through a third party before reaching the provider that runs the model.
  • Compliance is not demonstrated, with no GDPR mention, no DPA, no certification and no subprocessor list, which makes the service a poor place for sensitive or regulated personal data.
  • Economic and maturity risk go together here: prices and model availability can change, no given model is guaranteed to remain, rates may be revised thirty days after publication, and contractual documents left with template markers argue for caution before placing a production load on the service.
Setup

Setup & Integrations

Technical difficulty

Moderate, and squarely aimed at developers. You need an account, a generated API key and the habit of editing a base URL; the endpoint follows the OpenAI standard, so no code change is required, curl and PHP samples sit on the homepage, and each client tool has its own documentation page. The Web UI the site mentions is neither documented nor illustrated. The real friction lies elsewhere: choosing among 421 catalogue entries and decoding the Groups labels, such as ClaudeCode-Exc, Codex-Exc, Ent-HA or Pure-AZ, which no public page explains. Credits must be bought before any billed call.

Deployment

Web appAPI

Integrations

Claude Code Cursor Cline Codex OpenCode OpenClaw Gemini CLI Qwen Code

Supported languages

EnglishChineseArabicGermanSpanishFrenchJapaneseDutchPortugueseRussian
Company

Behind CodingPlanX

Company name
DeepNail Limited
Founded
08/05/2026
Country of origin
🇸🇬 Singapore
UBO
INFORMATION_NOT_FOUND
UBO country
INFORMATION_NOT_FOUND
Domain registrar country
🇺🇸 United States
Legal contact
Support contact

Social

Official links

Resources

All the official URLs gathered for verification and reference.

FAQ

Frequently asked questions

How many models does CodingPlanX actually give access to?
It depends which page you read, and the gap is wide. The homepage claims 600+ models, the documentation page states AI Models 100+, and the catalogue itself announced 421 models across multiple vendors and groups when it was consulted on 10 August 2026. The catalogue figure is the only one that can be counted.
Do I have to change my code to use it?
No. The service follows the OpenAI standard, so updating the base URL and the API key of an existing client is enough. Anthropic and Vertex protocols are also supported, and the site mentions a Web UI, which it neither documents nor shows.
Which tools are officially supported?
Claude Code, Cursor, Cline, Codex, OpenCode, OpenClaw, Gemini CLI and Qwen Code are the integrations put forward, and the documentation carries a dedicated page for most of them.
Is there a free plan or a free trial?
Neither is documented. No permanent free plan appears on the pricing page and no free trial is announced. The only way to obtain credits without paying is the referral programme, which grants USD 0.50 for a social share and USD 0.50 per friend who registers.
What does it cost to get started?
The cheapest paid option is the Lite pack, 10 credits for USD 10, with no commitment. Consumption is then billed per million tokens for text models or per call for image and video models, and taxes are not included.
Can I get a refund?
Yes, within 24 hours of a purchase or subscription, with a 5% platform fee retained and premium pay-as-you-go credits forfeited from the moment the request is made. Processing takes up to seven business days, followed by three to ten banking days.
Is my content used to train models?
The privacy policy states that the publisher may store, transfer or train on anonymised user content in order to improve its model. No option to exclude your data from that training is documented anywhere on the site.
Where is my data stored?
In Singapore, according to the privacy policy, which adds that personal information may be transferred or stored outside the user's country of residence, provided the recipient offers a comparable level of protection.
Is the service GDPR compliant?
The site never raises the question. The GDPR is not mentioned on any page collected, no EU representative is designated, no data protection officer is named and no DPA is published. The only rights described are access, update and correction, exercised by email.
Who runs the service, and which law applies?
The operator named in the terms is DeepNail Limited. Singapore law governs the terms and Singapore courts have exclusive jurisdiction. The minimum age to use the service is 18, or the local age of majority if it is higher.
Conclusion

Should you pick CodingPlanX?

The proposition is clear and genuinely useful: one key, one balance and one invoice for models drawn from many vendors, reachable over the OpenAI protocol and therefore without touching existing code. Prices are published model by model, and the advertised savings of 79% to 91% give a directly checkable reason to look.

The rest calls for a colder eye. The service is very young - domain registered on 7 March 2026, first Wayback capture on 8 May 2026 - and its publisher is barely documented: DeepNail Limited, no postal address, no named officer, no identifiable beneficial owner. The finish is uneven, and in one place contractually so: the card withholding agreement still carries unreplaced template markers, the English contact page ends with a Chinese translation note, the nine homepage testimonials bear template names, the blog is empty, and every page canonicalises to another domain. Even the model count is announced three different ways by the site itself.

Data governance is minimal rather than hostile: storage in Singapore, international transfers foreseen, training on anonymised user content with no documented opt-out, no DPA and no mention of the GDPR anywhere. The business model carries its own structural risk, since this is a reseller of access that depends on upstream providers and declines to guarantee that any given model stays available.

A reasonable use is therefore the one the product is actually built for: projects where inference cost is the binding constraint, where a model can be swapped without drama, and where nothing personal or regulated goes through the pipe. Where a compliance requirement, a procurement review or a long-lived production dependency is involved, the missing pieces decide the matter, not the price.