# Reported issues for playwright-report-mcp Pod holds 16 of 45 problems reported by people outside the maintainer team. Issues filed by the project's own owners, members and collaborators are excluded entirely — a maintainer's release checklist is not a warning to a prospective user. Back to [playwright-report-mcp](/mcp/playwright-report-mcp). ## Most discussed ### [ci] Extend MCP registry smoke-check retry window ## User Story As a tester, I want registry publication verification to tolerate indexing delays so that successful MCP releases are not reported as failed. ## Context The `v3.3.0` release exposed an eventual-consistency problem in `.github/workflows/publish-mcp.yml:88-105`. Run https://github.com/hubertgajewski/playwright-report-mcp/actions/runs/31577349819 successfully completed: - version synchronization - MCP publisher installation and authentication - `mcp-publisher publish` The subseq [Read the thread](https://github.com/hubertgajewski/playwright-report-mcp/issues/150) · 2026-08-12 · closed · 0 comments ### [chore] Bump to 3.3.0 for dual-era MCP protocol support # [chore] Bump to 3.3.0 for dual-era MCP protocol support ## Context PR #147 closed #146 and added backward-compatible support for both modern and legacy MCP protocol eras: - Modern clients can negotiate MCP revision `2026-07-28` through `server/discover`. - Existing clients can continue using the legacy 2025-era `initialize` lifecycle. - The selected protocol era is pinned for the connection lifetime. - Unsupported pinned revisions fail with an explicit negotiation error instead of silently [Read the thread](https://github.com/hubertgajewski/playwright-report-mcp/issues/148) · 2026-08-12 · closed · 0 comments ### [enhancement] Support MCP 2026-07-28 while preserving legacy clients ## User Story As a tester, I want `playwright-report-mcp` to negotiate either MCP `2026-07-28` or the legacy 2025-era protocol so that I can use modern stateless clients without breaking existing MCP integrations. ## Context The server currently depends on `@modelcontextprotocol/sdk` v1 in `package.json:43` and imports its server APIs in `index.ts:2-3`. The CLI constructs one global `McpServer` at `index.ts:621` and connects it directly to `StdioServerTransport` at `index.ts:962-965`. Direct [Read the thread](https://github.com/hubertgajewski/playwright-report-mcp/issues/146) · 2026-08-12 · closed · 0 comments ### [chore] Bump to 3.2.2 for split runtime modules # [chore] Bump to 3.2.2 for split runtime modules ## Context PR #120 closed #119 and split the monolithic implementation/test suite into focused modules while preserving public MCP server and CLI behavior: - Runtime source now lives under `src/`, with compiled output under `dist/`. - `src/index.ts` remains the executable entrypoint, and package `main` / `bin` still point at `dist/index.js`. - Package publishing now includes `dist/**` so all required compiled runtime modules ship with the [Read the thread](https://github.com/hubertgajewski/playwright-report-mcp/issues/121) · 2026-06-04 · closed · 0 comments ### [enhancement] Split monolithic implementation and tests into focused modules ## User Story As a maintainer, I want the monolithic implementation and test files split into focused modules so that future changes are easier to review, test, and publish safely as an npm package. ## Context The project currently keeps nearly all runtime code in `index.ts`: - `index.ts` is about 966 lines. - `test/server.test.ts` is about 2,466 lines. - `package.json` currently points both `main` and `bin` at `dist/index.js`. - `package.json` currently publishes only `dist/index.js` via `f [Read the thread](https://github.com/hubertgajewski/playwright-report-mcp/issues/119) · 2026-06-04 · closed · 0 comments ### [chore] Bump to 3.2.1 for token-efficient run status polling # [chore] Bump to 3.2.1 for token-efficient run status polling ## Context PR #115 fixed excessive token usage in the `v3.2.0` non-blocking Playwright run polling workflow: - `get_run_status` no longer returns retained `stdoutTail` and `stderrTail` text on every poll. - The server parses compact Playwright progress markers such as `[528/662]` from stdout without retaining raw stdout/stderr output. - `get_run_status` now returns `progress: { current, total }`, with nullable values before [Read the thread](https://github.com/hubertgajewski/playwright-report-mcp/issues/116) · 2026-05-29 · closed · 0 comments ### [bug] get_run_status polling consumes too many tokens ## Bug `get_run_status` in `v3.2.0` consumes too many tokens during normal polling of long-running Playwright suites because every status response includes retained stdout/stderr tails. ## Actual behavior `get_run_status` returns `stdoutTail` and `stderrTail` on every poll. Those tails are stored on the tracked run and capped only by `RUN_TAIL_LIMIT = 20_000` per stream in `index.ts`. In a full-suite measurement, 9 status polls produced 229,457 text chars from `get_run_status`; stdout/stderr [Read the thread](https://github.com/hubertgajewski/playwright-report-mcp/issues/114) · 2026-05-29 · closed · 0 comments ### [chore] Bump to 3.2.0 for non-blocking run status polling # [chore] Bump to 3.2.0 for non-blocking run status polling ## Context PR #111 added backward-compatible MCP API surface for long-running Playwright test workflows: - `run_tests({ wait: false })` starts a tracked background Playwright run and returns a stable `runId` immediately. - `get_run_status` reports active and terminal run state, output tails, `results.json` metadata, parsed stats, and idle status for a working directory when no run is tracked. PR: https://github.com/hubertgajew [Read the thread](https://github.com/hubertgajewski/playwright-report-mcp/issues/112) · 2026-05-29 · closed · 0 comments ## Most recent ### [enhancement] Add non-blocking test run status polling ## User Story As an AI coding agent, I want to start a Playwright test run without blocking and poll its current status so that I can understand whether tests are still running, failing, timing out, or producing report output during long runs. ## Context `run_tests` currently runs Playwright through synchronous `spawnSync` in `index.ts`, via `runPlaywright(...)`. While that child process is running, the MCP server process cannot serve another tool call from the same client. That means an agen [Read the thread](https://github.com/hubertgajewski/playwright-report-mcp/issues/110) · 2026-05-29 · closed · 0 comments ### [chore] Bump to 3.1.4 after Node 26 CI validation # [chore] Bump to 3.1.4 after Node 26 CI validation ## Context PR #103 added Node.js 26 to the CI test matrix and verified the package against Node 22, 24, and 26: - PR: https://github.com/hubertgajewski/playwright-report-mcp/pull/103 - Merge commit: `9fc8b09e15d7cbccd841f2d9245b1922f1ff2eea` - Issue: #102 The current release is `v3.1.3`, and the version fields still point at `3.1.3`: | File | Field | Current | Target | | --- | --- | --- | --- | | `package.json` | `version` | 3.1.3 | 3.1.4 [Read the thread](https://github.com/hubertgajewski/playwright-report-mcp/issues/104) · 2026-05-09 · closed · 0 comments ### [chore] Bump to 3.1.3 # [chore] Bump to 3.1.3 ## Context The latest published npm and GitHub release version is 3.1.2. The release workflow requires the package metadata and MCP registry metadata to stay in sync before cutting a new v* tag. This is a PATCH bump from 3.1.2 to 3.1.3 for the next release. ## Scope Update all version fields to 3.1.3 in one change: | File | Field | Current | Target | | --- | --- | --- | --- | | package.json | version | 3.1.2 | 3.1.3 | | package-lock.json | root version entries | 3.1 [Read the thread](https://github.com/hubertgajewski/playwright-report-mcp/issues/100) · 2026-05-09 · closed · 0 comments ### [chore] Bump to 3.1.2 for MCP publisher OIDC fix # [chore] Bump to 3.1.2 for MCP publisher OIDC fix ## Context The `v3.1.1` release completed successfully for GitHub Releases and npm, but the chained MCP registry publish failed during `Authenticate via GitHub OIDC`: - Release: https://github.com/hubertgajewski/playwright-report-mcp/releases/tag/v3.1.1 - Failed MCP publish run: https://github.com/hubertgajewski/playwright-report-mcp/actions/runs/25564862587 - Failure: registry expected OIDC audience `https://registry.modelcontextprotocol.io` [Read the thread](https://github.com/hubertgajewski/playwright-report-mcp/issues/97) · 2026-05-08 · closed · 0 comments ### [ci] Bump MCP publisher for registry OIDC audience # [ci] Bump MCP publisher for registry OIDC audience ## Context The `v3.1.1` release was cut from PR #94 and `release.yml` completed successfully: - GitHub Release: https://github.com/hubertgajewski/playwright-report-mcp/releases/tag/v3.1.1 - npm now reports `playwright-report-mcp@3.1.1` The chained `publish-mcp.yml` workflow failed before publishing to the MCP registry: - Failed run: https://github.com/hubertgajewski/playwright-report-mcp/actions/runs/25564862587 - Failure step: `Authentic [Read the thread](https://github.com/hubertgajewski/playwright-report-mcp/issues/95) · 2026-05-08 · closed · 0 comments ### [chore] Bump to 3.1.1 for dependency updates # [chore] Bump to 3.1.1 for dependency updates ## Context PR #90 updated the lockfile for `zod` from `4.3.6` to `4.4.3`. PR #91 updated the lockfile for `ip-address` from `10.1.0` to `10.2.0` and `express-rate-limit` from `8.3.1` to `8.5.1`. PR #92 updated the lockfile for `hono` from `4.12.14` to `4.12.18`, including upstream security fixes. These are dependency maintenance updates after the `3.1.0` release. There are no intended public API changes, tool schema changes, or behavior changes. [Read the thread](https://github.com/hubertgajewski/playwright-report-mcp/issues/93) · 2026-05-08 · closed · 0 comments ### [chore] Bump to 3.1.0 for run_tests CLI flag additions # [chore] Bump to 3.1.0 for run_tests CLI flag additions ## Context PR #85 (merged as `bfa7b3d`, closes #84) added 6 new optional input fields to `run_tests` — `updateSnapshots`, `headed`, `workers`, `retries`, `maxFailures`, `trace`. All changes are **additive** (new optional fields, no renames, no removals, no behavior change for callers who do not set the new fields). Under semver that's a **MINOR** bump from the current 3.0.0 to **3.1.0**: - MAJOR (4.0.0) — overstates it, nothing broke. [Read the thread](https://github.com/hubertgajewski/playwright-report-mcp/issues/86) · 2026-04-24 · closed · 0 comments ### [feat] Expose common Playwright CLI flags on run_tests (updateSnapshots, headed, workers, retries, maxFailures, trace) # [feat] Expose common Playwright CLI flags on `run_tests` (updateSnapshots, headed, workers, retries, maxFailures, trace) ## Context `run_tests` currently exposes only `spec`, `browser`, `tag`, and `timeout`. Consumers with workflows that should flow through the MCP — notably visual-regression baseline refreshes (`--update-snapshots`) — have to fall back to running `npx playwright test` directly. That workaround is especially painful in agent environments where a PreToolUse hook blocks direct [Read the thread](https://github.com/hubertgajewski/playwright-report-mcp/issues/84) · 2026-04-24 · closed · 0 comments The remaining reports are on [the project's issue tracker](https://github.com/hubertgajewski/playwright-report-mcp/issues).