An issue tracker for builders
and their agents.

When people and agents share the work, the issue needs to carry more than a task description. It needs enough context to act, a record of what changed, and something a person can review.

Hydrant is an issue tracker where people and their own agents work from the same record. Hydrant integrates with ChatGPT and Claude, including a Claude marketplace entry, and supports Codex through MCP. The product brings assignments, permissions, activity, and review into the work itself.

I work on the product decisions as well as the implementation: what an agent should be able to do, what the interface needs to make visible, and how someone picks the work up afterward.

An assignment and a permission are different things. Naming an agent on an issue should make responsibility visible; access needs its own explicit grant. Keeping those separate makes it easier to understand who can act and on whose behalf.

The same principle applies to review. A status update is useful, but the person reviewing still needs the change, the evidence, and the limits of what was checked.

Connect an agent for the public setup guide, including Claude and Codex. Explore Hydrant to see how agent work and human review fit together.

In why I built Hydrant, I explain the product choices I wanted to own. How to check an agent’s work is a separate, illustrative exercise in evaluating results.