Hillclimb replaces your ticket backlog with a living definition of what your software must do. People and agents work through ranked outcomes, building what's missing and verifying it with evidence.
As requirements change, the work follows.
See it in practicePaste these instructions into your agent to get started.
Read https://hillclimb.me/llms.txt and connect to Hillclimb. Help me define what my software must do as ranked, checkable outcomes. Once I approve them, loop through the tree: build, verify with evidence, and pick the next outcome.
Claude Code, Codex, Cursor, or your agent of choice.
A missed reminder becomes a fix, evidence, and a verified outcome. New evidence brings it back for another pass.
Hillclimb puts the broken reminder outcome first, in the order set by the business.
Explore a fictional scheduling product. Each outcome defines a behavior to build and verify, connected to the goal it serves.
Define the behaviors your system must support, how to check them, and which matter most. That tree becomes the shared reference for everyone building the software.
You approve the requirements. Your people and agents keep climbing against them, from the first implementation through every change that follows.
Read the agent instructionsTake the highest-priority outcome that needs work. Read its requirements and the goals above it.
Implement or fix the underlying software to meet the agreed criteria.
Check the actual behavior and record the results. Mark the outcome verified when its criteria hold.
Move to the next outcome. Changed requirements and new failures bring outcomes back for another pass.
See which requirements are verified, which need another check, and what to build next.