Player profile
Kraft
An Apache-2.0 local orchestrator that runs each task from spec to pull request in its own git worktree, so tasks run in parallel.
Profile
Where it fits
Kraft is an Apache-2.0 local orchestrator that runs each task from spec to pull request in its own git worktree, so tasks run in parallel. Steps run in order from a YAML chain file, and tests and CI are checked in steps of their own.
Attempts and spend are capped. A person approves the spec, the plan and the final review from the web UI, a phone or the CLI. Claude Code by default; Codex, Cursor, Gemini CLI, OpenCode, Amp and Antigravity can run steps.
Builder web analytics
Measure outcomes from Kraft-run agent work
Kraft runs each task from spec to pull request in its own git worktree. Separately configured product analytics can show the observed user-facing results of a change a person approves.
This is an optional measurement workflow, not a native Kraft integration with Agent Analytics. Configure product tracking and authorized analytics access separately; the listing does not install tracking or report Kraft task activity.
First loop to measure
- Kraft runs a task through a YAML chain in its own git worktree, and a person approves the spec, the plan and the final review
- the reviewed change is deployed to a separately instrumented project
- a human or an authorized agent reads observed product outcomes
- the team uses that evidence to choose the next task
Copyable prompt
If product tracking and analytics access are already configured, inspect the deployed change over an appropriate observation window. Report measured outcomes, missing evidence and limitations. Do not infer causation from before-and-after movement or treat task activity as product usage.