[ Coding agents ]

Playwright CI failures, in your coding agent

kinora ships an MCP server. Point Claude Code, Cursor, or any MCP client at it and your agent pulls the failing test, its trace, and flaky history straight into the chat, then works the fix. Local over stdio, against the cloud or your own self-host.

Get started Live demo

Copy-pasting a stack trace into your agent was step one. kinora exposes your CI test data over the Model Context Protocol, so the agent reads the last run's failures, opens the Playwright trace, and checks whether a test is a fresh regression or a chronic flake, without you shuttling files around. It runs locally over stdio against kinora cloud or your self-host, scoped to your API token.

mcp config

{
  "mcpServers": {
    "kinora": {
      "command": "npx",
      "args": ["-y", "@kinora/mcp"],
      "env": { "KINORA_TOKEN": "<token>" }
    }
  }
}

list_failures the last run's failures, with error and file:line
get_trace the Playwright trace for any failing test
test_history flaky-vs-regression history per test
list_projects every project and its latest run
get_run the full report for a single run

Five tools, zero glue

The agent calls the tools directly over MCP, no scraping and no copy-paste.

Straight into the chat

Point Claude Code, Cursor, or any MCP client at kinora and the failing test, its error and file:line, and the trace land in context.

Regression or flake

test_history gives the agent pass/fail/flaky rates, so it fixes a real regression instead of chasing a chronic flake.

Local, cloud, or self-host

Runs over stdio with your API token, against kinora cloud or your own self-hosted server. Your data stays yours.

FAQ

How do I connect a coding agent to my Playwright tests?

Add the kinora MCP server (@kinora/mcp) to your MCP client with a kinora API token. Claude Code, Cursor, or any MCP client can then pull failures, traces, and history from CI.

Which MCP tools does kinora expose?

Five: list_failures, get_trace, test_history, list_projects and get_run, covering the last run's failures with error and file:line, the trace for any failing test, per-test flaky-vs-regression history, every project with its latest run, and the full report for a single run.

Does it work with self-hosted kinora?

Yes. The MCP server runs locally over stdio and points at either kinora cloud or your self-hosted server, authenticated with your API token.

Awaiting first signal

Push your first run.

Add the reporter, point it at a kinora server, and watch the history build itself.

Get started Live demo Star on GitHub