XHawk 0.99 - Transform Coding Sessions & Code into a System of Context
by•
Turn your session history into a knowledge base. XHawk CLI automatically captures AI sessions during every git push, mapping the agent's reasoning directly to your commits. Don't just ship code. Capture the intent, audit the logic, and build a searchable, collaborative record of how your software actually gets built with coding agents. Our AI decodes your entire codebase, generating dynamic learning paths and docs for agents and humans.
Replies
Congratulations
OS Ninja
@madalina_barbu Thanks for your support!
OS Ninja
Huge shoutout to our teams spread across California and the Indian sub-continent! Thanks everyone.
Congratulations on the launch! If you analyze the codebase, can you also give improvement suggestions? I think that would be very useful.
OS Ninja
@mykyta_semenov_ Thanks for your feedback. We have a rich roadmap. Some of these features will be built based on the requests coming in from the customers
We manually maintain CLAUDE.md across our whole monorepo, it's a real struggle. When building multi-agent architecture, especially when embedding context or passing feedback, everything just gets lost and you can't restore the session. So this looks like a pretty interesting idea.
Used it and loved it!
OS Ninja
@aaditya_menon Thanks for your support!
Congrats on the launch, punit!
OS Ninja
@akhilbvs Thanks!
Congrats on the launch! The idea of capturing the reasoning behind code, not just the code itself, is something engineering teams have needed for a long time. Most knowledge walks out the door when a developer leaves or a context window resets.
The AGENTS.md standard is an interesting move. Would be curious to know how it handles projects where the architecture has evolved a lot over time, and the early decisions are basically archaeology at this point.
OS Ninja
@andreitudor14 We update the AGENTS.md files every night based on code changes. The key idea is that we should capture the patterns and anti-patterns from the codebase automatically. Every developer should not start from scratch. The biggest challenge is that LLMs are stateless. Engineering teams are successful because they are stateful. Our goal is to bring that "collective intelligence" into the software development lifecycle. It will be mostly driven by agents in the future so we are optimizing for that.
The idea of making AI reasoning auditable is really compelling — most of the value in a coding session lives in why a decision was made, not just what was committed. Curious how it handles sessions where the reasoning was messy or exploratory (lots of dead ends) — does it try to clean that up into a coherent narrative, or preserve the full chaos?
OS Ninja
@echosun thanks for your feedback. We have built some IP to track relevance and reduce noise. It's a hard problem since a single commit can have multiple sessions. Take a look at our UI - we provide summary view and also index the sessions so that they can be queried later.
@puneet_singh25 That's a smart approach — surfacing a summary view rather than dumping raw session data is the right call. One follow-up: does the summary get generated at commit time, or can you trigger it manually mid-session when you want to checkpoint your thinking?
OS Ninja
@echosun It's flexible and we will make it configurable so that users have control. For now, we are doing it at commit time in the 0.99v release. Thanks again for your insightful questions!
@puneet_singh25 That makes sense — the summary view approach is smart, it abstracts away the noise without losing the intent. One thing I'm curious about: does the relevance tracking work across branches? Like if I'm context-switching between features, does it keep those session threads separate, or does it blend them together?