Haystack
Haystack provides automated QA for AI coding agents by replaying real customer journeys against every code change. The platform builds the app twice—with and without the change—and explores both side by side on isolated machines to pinpoint behavioral differences. Results typically arrive in 10 to 15 minutes via the haystack verify CLI.
- Developer Tools
- Software Only
Funding
Raised to date
$500KRaised to date based on public sources. This may differ from the amount the company actually raised and is based only on what is publicly available on the internet.
Founders
Founder details are not available yet.
Product
Problem
Traditional test suites often miss silent bugs because they rely on manually scripted test cases that do not reflect real customer journeys. When coding agents make changes, developers lack a fast, automated way to see exactly what behavior broke across realistic scenarios, leading to regressions that surface only after deployment.
Solution
Haystack provides automated QA by replaying real customer journeys against every code change. It crawls the application by building it twice—once at the base commit and once with the change—and explores outward from the changed code, comparing both copies side by side to identify behavioral differences. Each copy runs on Haystack's isolated machines with stand-in data and test accounts created during onboarding, so nothing touches production systems. The CLI, `haystack verify`, names the changed files, waits for the crawl, and prints each step and each bug as it is found, typically delivering an answer in 10 to 15 minutes.
Target Audience
Primary users are software developers and engineering teams using AI coding agents who need fast, automated verification that code changes do not break real customer workflows.
Features
- No configuration file: works out how to run the app from existing lockfiles, Dockerfiles, compose files, and package scripts
- Isolated machine copies with no internet access and stand-in data, ensuring only the change under test differs
- Side-by-side exploration with automated judging of every difference to pinpoint regressions
- CLI options including `--minutes` (1-30, default 3), `--json` for a single JSON report, `--raw` for the full crawl record, and `--no-wait` to return current state
- Exit codes distinguishing success, failure, crawl interruption, blocked onboarding, handshake errors, and broken app copies
- Integration with coding agents via a Claude Code stop hook and an AGENTS.md note, with a setup command for Claude Code, Codex, or Cursor