Launching today
Oxagen
Mission control for agents to share + maintain context
47 followers
Mission control for agents to share + maintain context
47 followers
Oxagen links every coding-agent run to the people, work, code changes, costs, and outcomes behind it. Engineering teams use that record to direct their agents, share proven instructions and tools, enforce policy on the actions agents take through Oxagen, and get better at the next task. We built it so teams can see real returns on their AI spend, and so great engineers can still do excellent work in the age of agents.
Free Options
Launch Team
LoomTalk it through in Loom. Ship with AI
Promoted
Hi Product Hunt, I'm Mac, and I built Oxagen.
If your team runs Claude Code, Codex, and Cursor on the same repositories, you know how the week ends: dozens of agent runs, a model bill nobody can split by person or by work item, and no one place that says which run changed which line.
Oxagen connects each coding-agent run to the person who started it, the work item behind it, the code it changed, what it cost, and how it ended. Engineering teams use that record to:
steer supported agents with steering their team reviewed
share toolbelts and skills across Claude Code, Codex, Cursor, and Stella
enforce budgets and approval rules on the actions routed through Oxagen
see spend by operator, agent, and workspace
Get started in 60 seconds or less - no browser or credit card required:
Navigate to https://docs.oxagen.sh/ to get started today. No credit card required.
How do you run coding agents today, and where does it break?
Nice launch. When several agents update the same piece of context, how do you settle conflicts? Is there versioning or an audit trail so we can see which agent changed what?
@zain_sheikh context is managed in a gitops style way, each record is saved as a file in a steering repo in git making version control inherent in the underlying tech. "Context PR" is a thing in Oxagen and it also gives teams the benefit of being able to integrate a review process around shared steering.
The record-keeping is the part that got me — I run Claude Code and Cursor on the same repo, and by the end of the week I can't tell which run changed what, or what it cost. Tying each run to the person, the work item, and the diff is exactly the record I've been missing. One thing I'm curious about: does that record only cover runs routed through Oxagen, or can it pick up runs I started directly in Claude Code?
@jcbuilds You can pickup any run on Claude Code or Codex or Cursor. You can backfill runs also. And depending on your settings you can ingest the full transcript or do it in digest mode only. Every prompt, every tool call, every policy decision recorded along with the operator of the agent and the runtime/device.
Does the reviewer get to see the prompt a run started from?
@leoheo 100% yes everything is recorded. Oxagen is more of an active tool where engineers go to review work assigned to them and dispatch agents to complete that work. I encourage you to go to the demo site which runs read only data (fixtures) on the real app surface. https://demo.oxagen.app