Skip to main content
Hopper is the built-in AI chat assistant powered by a highly optimized Retrieval-Augmented Generation (RAG) pipeline. Unlike generic AI assistants, Hopper knows your project’s rules.

​
Core Capabilities

When you ask a question, Hopper converts your query into a vector embedding and searches your active Capsule for the most relevant decisions and memories.
Before streaming a response, Hopper scans its own generated output. If it detects a violation against an ACCEPTED decision, it attaches warning badges to the response.
Manually tag specific decisions or memories in the chat input to force Hopper to focus solely on those constraints.

​
AI without the Hallucinations

Hopper operates strictly as an advisory layer. It can explain why OAuth 2.0 was chosen by reading your memories, and it can generate code snippets based on that rule, but it cannot override the authority of the decision-api. You get real-time, streaming responses powered by OpenAI, reranked for 95% relevance improvement using Cohere.