AgentSpend finds repeated repo work, highlights avoidable model spend, and produces a before/after savings report.
AgentSpend analyzes coding-agent usage, identifies repeated repo work, and turns the result into a savings report before teams change workflow.
The first step is a savings audit. Developers keep using their existing tools while AgentSpend identifies where repeated work creates waste.
Reduction starts only on a controlled workflow with a quality metric the team agrees to before rollout.
Each pilot ends with a clear report: repeated work found, estimated avoidable spend, assumptions, and quality result.
AgentSpend creates a compact repo pack, agent instructions, and a savings report. Your coding agent reads the pack first instead of repeatedly rediscovering the same project structure.
# one command inside your repo cd your-repo npx -y https://18.207.249.172.sslip.io/agentspend-cli-0.6.2.tgz start --email you@example.com # or start and run your coding agent npx -y https://18.207.249.172.sslip.io/agentspend-cli-0.6.2.tgz start --email you@example.com -- codex # give your coding agent: Read .agentspend/repo-pack.md first.
Private-alpha testing found a strong repeated-work signal in real coding-agent usage. The next milestone is pilot validation against customer usage and invoice data.
The CLI report focuses on the facts a buyer needs: indexed files, compact context size, repeated broad-read estimate, avoidable tokens, and cost model assumptions.
AgentSpend scan complete Repo pack: .agentspend/repo-pack.md Report: .agentspend/savings-report.md Estimated tokens avoided per repeated task: 10,663 Estimated saved per 100 repeated tasks: $2.13 next step: use the repo pack in your coding-agent workflow
Start with the CLI. If the report shows material repeated context waste, run a one-week pilot against real provider usage and optimize one controlled workflow.