--- title: "xAI REST API Review: Developer Experience, Docs & Pricing Rated" description: "A hands-on review of the xAI REST API covering onboarding, documentation quality, developer portal features, and a final score of 19/30. See how xAI compares." canonicalUrl: "https://zuplo.com/learning-center/xai-rest-api-review" pageType: "learning-center" authors: "adrian" tags: "API Best Practices, APIs" image: "https://zuplo.com/og?text=xAI%20REST%20API%20Review%3A%20DX%2C%20Docs%20%26%20Pricing" --- ## Introducing the xAI API It's always an exciting day when a major tech company launches their first API, and the internet was certainly buzzing when Elon announced the new [xAI API](https://docs.x.ai/docs). ![xAI API release tweet](/media/posts/2024-10-21-xai-rest-api-review/image-1.png) Most AI platform companies (ex. OpenAI and Anthropic) see a large amount of traffic from companies integrating their REST APIs to add AI features to their apps/websites. [xAI](https://x.ai/) has long been seen as an underdog in the AI race, however with a [recent funding round of $6B](https://x.ai/news/series-b), the platform has accelerated development to catch up with its rivals. The question that lingered on my mind is if the developer experience of xAI might be good enough to lure away users of OpenAI or Anthropic's APIs. ## xAI API Functionality The xAI Rest API, powered by Grok, provides a robust set of capabilities for developers to build AI-driven applications. The API grants access to [Grok](https://x.ai/news/grok-2), a powerful large language model designed for various tasks, including text and code generation, as well as image-related functionalities. Key capabilities include: - **Text and Code**: The API can generate text and code, extract data, create summaries, and support various natural language processing tasks - **Vision**: The API supports image generation, object identification, and visual analysis, which can be useful in applications that require visual data interpretation or content creation. - **Function Calling**: It offers advanced function-calling capabilities, allowing developers to connect the model to external tools, which enables handling complex workflows involving external data and functions. - **Model Options**: Developers can choose between Grok-2, which is optimized for complex reasoning tasks, and Grok-2 mini, a lighter model offering faster responses with a trade-off in output depth. The API also supports streaming responses for real-time applications like chatbots. It’s compatible with the [OpenAI SDK](https://docs.x.ai/docs/guides/migration), making it easier for developers familiar with that platform to integrate xAI with minimal adjustments. (Note: Anthropic SDK compatibility, which was available at launch, has since been deprecated.) For a deeper look at integrating the xAI API into production systems, see our [xAI API integration guide](/learning-center/xai-grok-api). ## Onboarding Experience Getting started with the xAI REST API is quite straightforward. Simply create an [xAI account](https://accounts.x.ai). You can choose to authenticate with your X account, Google, Microsoft, or plain-old email + password. ![xAI login](/media/posts/2024-10-21-xai-rest-api-review/image-3.png) From what I observed, there's a different onboarding for users coming from free-mails (ex. gmail) compared to work email domains. The work-mail onboarding flow guides you through creating an account, inviting team members, and importantly - getting your first API key. ![xAI Company Signup](/media/posts/2024-10-21-xai-rest-api-review/image-5.png) The free-mail version dumps you out to your account management screen with a small button to go to the API console. Once you click the "API console" button, it takes you to the same flow as above. ![xAI Account Management](/media/posts/2024-10-21-xai-rest-api-review/image-7.png) ### Getting Your API Key & Time To First Call After signing up, you are prompted to create your first key, which is provisioned immediately, and accompanied by a `curl` code sample. ![xAI API key onboarding](/media/posts/2024-10-21-xai-rest-api-review/image-6.png) I really appreciate how quickly they aim to get users onboarded, this significantly reduces the [time-to-first-call](/learning-center/optimize-ttfc-with-api-keys) (TTFC), especially since you just need an API key rather than a complicated JWT setup. If you're weighing different approaches, our guide on [API authentication best practices](/learning-center/api-authentication) covers the trade-offs between API keys, OAuth, and JWTs. Other than the weird flow for free-mail users, I was feeling really good about the onboarding flow, until I actually tried making the API call. I was immediately greeted by an error: ```json { "code": "The caller does not have permission to execute the specified operation", "error": "Your new team 9e287ee1-ad55-41d6-91b4-24fe8ba21b40 has just been created and is still being initialized. Please try again shortly." } ``` I navigated to the API console to see what was going on (I thought the key was provisioned already?) and saw this banner up-top. ![team is initializing](/media/posts/2024-10-21-xai-rest-api-review/image-8.png) How perplexing - what "initialization" needs to be made here? Their API Key service seems to have already worked, and their models are already deployed - so what is going on behind the scenes? Disappointingly, this first call is also your last free one - you are forced to pay for the API after receiving that initial error. ![out of credits](/media/posts/2024-10-21-xai-rest-api-review/image-9.png) This feels like a bit of a bait & switch, the onboarding made me feel like there was some sort of free-tier, but instead, I am forced to pay to get started. The onboarding was very seamless up to this point, so despite this rug-pull, it still deserves a respectable score. **Onboarding Score: 7/10** ## xAI API Documentation Quality Now let's move onto the API docs. They are split between the [introductory materials](https://docs.x.ai/docs) and the [API Reference](https://docs.x.ai/docs/api-reference) which is common. The quickstart is just a more in-depth version of the onboarding flow, while the API reference goes into more depth around topics like authentication and streaming. The API reference is nicely laid out in the typical 2 column + navbar format. ![xAI Rest Endpoint example](/media/posts/2024-10-21-xai-rest-api-review/image-10.png) Here are my pros and cons **Pros** - Nice layout - Docs cover common usecases (ex. streaming) - All query parameters, including optional fields are documented - Some error codes are documented (ex. 400, 404) **Cons** - Many parameters lack descriptions so you have to infer what they mean from the examples provided - Definition and Example tab are separate for no reason - Headers needed are not documented inline - Naming doesn't follow a consistent convention (ex. `Completions` vs `Create embeddings`) - No interactive playground - Many errors are undocumented or poorly documented ### xAI OpenAPI Specification Does xAI offer an OpenAPI/Swagger spec? Initially, the answer was no because I couldn't find a link to it anywhere. I did notice that their docs were bespoke however, so I did some network tab sniffing and found it! The OpenAPI specification lives at [https://docs.x.ai/openapi.json](https://docs.x.ai/openapi.json). In its raw form — the OpenAPI spec is not pretty. If you want to learn how an OpenAPI spec can power great documentation, read our guide on [generating API documentation from OpenAPI](/learning-center/generate-api-documentation-openapi). I ran it through [RateMyOpenAPI](https://ratemyopenapi.com/report/3186af9f-b0ac-4f8f-b400-8ba3966ebe9b), and wasn't surprised by the low score. ![RateMyOpenAPI Score](/media/posts/2024-10-21-xai-rest-api-review/image-11.png) The lack of interactive playground in the docs bothered me a lot, so I decided to create my own. I had to modify the OpenAPI from its original state to make it compliant with most tooling. Here's the [end result](https://zudoku.dev/demo?api-url=https%3A%2F%2Fraw.githubusercontent.com%2Fzuplo-samples%2FxAI-OpenAPI%2Frefs%2Fheads%2Fmain%2FxAI-openapi.json). You can find the full OpenAPI specification [in this repository](https://github.com/zuplo-samples/xAI-OpenAPI/tree/main?tab=readme-ov-file). ### xAI Code Samples & SDKs Unfortunately, the API Reference does not provide any code samples (not even curl) beyond the onboarding sample. My [Zudoku project](https://zudoku.dev/demo?api-url=https://raw.githubusercontent.com/zuplo-samples/xAI-OpenAPI/refs/heads/main/xAI-openapi.json) above does include code samples in case you want them. The [integration docs](https://docs.x.ai/docs/guides/migration) make reference to a gRPC and Python SDK - but do not link to them. You can use the Grok API via the OpenAI SDK - which makes migration easier, but isn't useful for developers that aren't already using it. Overall, I feel like this documentation is below average, especially compared to other AI platform companies that offer richer docs, playgrounds, and SDKs. If you're curious about what separates good API docs from bad ones, check out our guide on [must-have features for API documentation tools](/learning-center/top-api-documentation-tool-features). **API Documentation Score: 3/10** ## xAI Developer Portal Features The developer portal for the xAI API is actually quite robust in its feature set. I suspect some engineers from X (formerly known as Twitter) were "loaned" to help with this. An engineer named Toby Phalen provided a [feature overview](https://x.com/TobyPhln/status/1848430127950270957) via X, covering some prominent features of the dev portal. ### xAI API Usage Analytics The API features a usage explorer which includes breakdowns by tokens and spend. ![xAI API Analytics](/media/posts/2024-10-21-xai-rest-api-review/image-12.png) ### xAI Billing & Pricing Speaking of spend, you can easily connect a credit card to your account to start purchasing credits. Pricing is determined by your choice of model. For example, the `grok-beta` model costs $5 per million input text tokens, and $15 per million output text tokens. ![model pricing](/media/posts/2024-10-21-xai-rest-api-review/image-15.png) There isn't a spend estimate tool to help you understand the translation from dollars to tokens - so you need to do your own math. If you're designing your own API's pricing model, our breakdown of [API pricing strategies](/learning-center/api-pricing-strategies) covers tiered, usage-based, and hybrid approaches. ### xAI User Management User management is quite simple, invite your coworkers by email and they will get access. There is also a nice feature where you can verify a domain and everyone who signs up with that domain will get access to your team. ![domain verification](/media/posts/2024-10-21-xai-rest-api-review/image-13.png) ### xAI Auth & Security As stated earlier, we are a big fan of using [API Keys](/learning-center/rebuttal-api-keys-can-do-everything) for API authentication, and creating/managing keys is quite simple in the xAI console. Account security is also top notch, with built-in session management and options for multi-factor authentication. ![2FA](/media/posts/2024-10-21-xai-rest-api-review/image-14.png) All things considered, this developer portal has everything you would expect from a modern API, and some nice security features. For a deeper look at what makes a great developer portal, see our guide on [what a developer portal is and why it matters](/learning-center/what-is-a-developer-portal). **Developer Portal Score: 9/10** ## xAI API Final Review Score: 19/30 Overall I'd give this API a **final score of 19/30** which is slightly above average for an API - but I definitely expected more given how long it took to release. Some obvious improvements to be made include: - Free credits to get started - Fully documented API docs with a playground and code samples - First-party SDKs rather than reusing OpenAI's It's a bit frustrating to give this API a low score, because it's _so close_ to greatness, but falls flat right at the end. ## Should You Consider the xAI API? The only reason to use the xAI API would be to get access to Grok, which differs greatly from GPT or Claude in terms of moderation and safety. So in reality, this API is likely your only option to use Grok. If you aren't too fussy about the model, and want a better DevX, I would recommend using OpenAI or Anthropic. ## Alternatives to the xAI REST API The only alternative I was able to find was via the [OpenRouter platform](https://openrouter.ai/x-ai/grok-beta). ## Building a Better API DevX Than xAI If you are revamping your API - and want to have: - Automatically issued API Keys that _actually work_ - Autogenerated API documentation with code samples and a live playground - Monetization system where you can offer a free tier - Live usage analytics for yourself and your users Then you need Zuplo - [get in touch](https://zuplo.com/meeting?utm_source=blog)!