Sniff
Run an MCP-connected QA and accessibility pass that walks real user flows against your live local or deployed app and reports breakages.
Overview
Sniff is an MCP server for the Ship phase that scans QA and accessibility by walking real user flows on your running app.
What is this MCP server?
- QA plus accessibility scanner oriented around real user flows on a running application
- Surfaces what is broken in flows agents might otherwise assume work
- GitHub-hosted MCP server (Aboudjem/sniff) at manifest version 0.7.1
- Pairs with pre-launch checks when solo builders skip dedicated QA headcount
- Agent-invokable via MCP instead of ad-hoc manual click paths
- Published server version 0.7.1 in MCP manifest
- Source repository github.com/Aboudjem/sniff
- Described scope: QA plus a11y on real user flows
Community signal: 4 GitHub stars.
What problem does it solve?
Solo shippers often miss broken flows and a11y issues because manual clicking every path does not scale before launch.
Who is it for?
Indie developers preparing a release who already have a runnable app and want agent-driven regression and a11y signal on critical journeys.
Skip if: Greenfield repos with no running instance, or teams needing formal compliance audit trails without reviewing Sniff output.
What do I get? / Deliverables
After wiring Sniff via MCP, your agent can trigger flow walks against the live app and get a concrete list of what is broken or inaccessible.
- Reports of broken or incomplete user flows on the running app
- Accessibility-oriented findings tied to traversed paths
- Actionable issue lists your agent can turn into fix tasks
Recommended MCP Servers
Journey fit
Sniff belongs on the Ship shelf because it validates runnable product behavior and a11y before you treat a release as customer-ready. Testing is the canonical subphase: it exercises end-user flows in a browser-like walkthrough rather than static code review or infra monitoring alone.
How it compares
Browser flow QA MCP scanner, not a unit-test framework or production APM dashboard.
Common Questions / FAQ
Who is Sniff for?
Solo builders and small teams using MCP agents who need quick validation that key user flows and accessibility basics work on a running build.
When should I use Sniff?
Use it in Ship/testing before launch or after meaningful UI changes—when the app is up and you want flow-level breakage and a11y findings, not during pure ideation.
How do I add Sniff to my agent?
Clone or install from github.com/Aboudjem/sniff, register the MCP server per repo instructions, point it at your running app URL, and invoke scan tools from Claude Code, Cursor, or compatible clients.