We're currently in private early access. If you're interested, please fill out the request access form or schedule a demo.

Yup. Inkeep's search works by applying large language models to understand your content and user queries. This allows us to give best-in-class results based on user intent, not just string matching. In addition, developers can engage with a ChatGPT-powered chat mode to ask questions and self-troubleshoot, right from the search bar.

You can add call to actions so that developers engage with you in whichever way you prefer (1:1 chat, Slack/Discord, email, etc.). Our AI chat option is designed to not get in the way and be authentic. Developers can optionally engage with it to self-troubleshoot and ideate, just like they do with GitHub copilot or ChatGPT. But, you can also turn it off and serve just search results if you prefer.

While chat responses might not always be 100% accurate, we've built our system to mitigate hallucinations and protect your brand. For example, we deprioritize old content, tailor the chat responses to be positive and helpful, and put guard rails in place so it only answers questions about your product. In addition, our admin dashboard gives you the ability to provide revised answers, choose between variations of answers, deprioritize search results, and more. With that and end-user feedback, the chat and search services become even better over time.

You can add a JS script to any existing site or page. We also make the chat and search bar available as React components for frameworks like Remix and Next.js.

You can use just the search bar if you'd like; however, you can also use the chat widget in place of Intercom, Drift, or other similar services. Unlike those services, we leverage all of your existing content with zero effort and tailor-make the experience for developer products. We suggest related content from your unified knowledge base, provide a ChatGPT-powered self-serve option, and make it easy for developers to get help in your preferred support channels -- without any fluff.

We've specialized in knowing how to index, chunk, and embed content from each integration we provide. This is crucial in order to get the best results, so we do it all for you including for your documentation. We have pipelines and listeners in place to keep the content up to date on a regular basis without any effort.

We use our own custom embeddings and large-language models specifically made for answering questions about technical products. The base layers include the OpenAI (like GPT4) and other open-source models. We use user & admin-generated feedback as well as details about your company to tailor our model to your products.