Skip to main content
Bugspray is an integrated QA system for developers and AI agents. It connects product expectations, exploratory QA, reproducible findings and automated validation so you can understand both what was checked and what the evidence supports.

Try it locally

Start the app and store your first synthetic finding. No provider account required.

Connect an AI agent

Give your agent the reporting guide, catalog and evidence tools through MCP.

​
Follow one behavior through the system

Imagine a small project app called Example Desk. A user saves a filtered view, reloads the page and expects the same filters to return.
  1. Describe the behavior. Create the product, its Views area and a “Save and reopen a view” workflow with a clear expected result.
  2. Investigate. A developer, QA reviewer or agent tests the behavior. When it breaks, record exact steps, observed impact and inspected screenshots or video.
  3. Reproduce and verify. Attach a deterministic Playwright repro and, on a trusted execution-enabled deployment, replay it against the original app or a candidate fix.
  4. Connect existing automation. A repository manifest declares which suites intend to cover that workflow. PR planning and result reporting retain the candidate, saved expectations and environment identity.
A workflow is an expectation. A finding is an observation. A suite declaration is intended coverage. A passing result requires actual execution evidence.

​
Choose your next step

Build your product catalog

Organize products, areas and durable workflow expectations.

Review findings and replay

Understand evidence, repro outcomes and candidate verification.

Validate a pull request

Read plans, exact identities, suite receipts and explicit gaps.

Run your own instance

Choose local Node or an operator-managed deployment.
These docs describe a project preparing for a future public source release. Repository access is currently required. Project licensing and bundled dependency redistribution remain separate owner decisions. Local use is available from an authorized checkout; no public hosted app demo is assumed here.

​
What is ready to use?

The local app supports findings, evidence, collaboration, the product catalog, suite inventory, HTTP and MCP. Browser authentication and repro execution need additional local dependencies and trusted test accounts. GitHub, Linear and Slack are optional connections. Suite planning and CI integration are experimental. A narrow three-test pilot has passed unattended against a verified preview with normal authentication and real API access. That does not establish generic automatic label-to-plan-to-CI-to-report operation, safe test skipping or auto-merge. See the PR validation guide for the boundaries.