v2.0.1

TestDino MCP v2.0.1

New package name: this release publishes as @testdino/mcp. If you were on the old testdino-mcp package, update your MCP config to @testdino/mcp (for example "args": ["-y", "@testdino/mcp"]). Same tools and behavior, new name.

TestDino MCP is no longer just about managing and analysing your tests. It now acts on them. From a failed test you can file a tracked bug in your issue tracker, cut through a big red run by grouping failures into real problems, and dig into test history with far more precision. All from natural language, right inside Cursor, Claude Code, or VS Code.

Turn failing tests into tracked issues, without leaving your editor

Connect your team's issue tracker once, then ask your AI assistant to file a ticket straight from a failed test or run. Supported trackers:

• Jira
• Linear
• Asana
• monday.com

What you can now do:

• Connect a tracker by asking your assistant to link the integration (it hands you a secure authorization link to approve in your browser).
• File an issue from a failure with the test context already attached, so the ticket lands with the right details instead of a blank template.
• Check issue status later by asking for updates on tickets you have already filed, including their current state in the external tool. Ask about several at once.
• See what's connected for a project before you start, so you always know which trackers are live.

The result: the gap between "a test failed" and "a ticket exists for it" is now one sentence.

Triage big failing runs by pattern, not one test at a time

When a run fails with dozens of red tests, you no longer have to read them one by one. Ask your assistant to cluster the failures and it groups tests that share the same underlying error. You immediately see which failures are one real problem versus many, so you can fix the root cause first and stop chasing duplicates.

Sharper filtering and history for test runs and cases

Finding the exact run or test you care about is faster now:

• Filter runs by status, by test case tags, by free-text search, and sort the results, so you can zero in on the runs that matter instead of scrolling.
• Filter for artifacts to find tests that actually have screenshots, videos, or traces available, including artifacts from earlier attempts of a flaky test.
• Pull test history and step detail for a case, so you can see how a test has behaved over time and what happened at each step.

Know your access at a glance

The health check now shows your role in the organization, so you can confirm what you are able to do before you try it.

New tools in this release

Five new capabilities your AI assistant can now call:

• connect_integration — link an issue tracker (Jira, Linear, Asana, monday.com) to a project.
• get_integration_status — check which trackers are connected before you act.
• create_external_issue — file a ticket from a failing test or run.
• get_external_issue — pull the current status of tickets you have filed.
• get_run_error_clusters — group a run's failures by their underlying error to triage at scale.

Notes for existing users

A few things changed that need a one-time update:

• New package name. Point your MCP config at @testdino/mcp instead of testdino-mcp. The old package is no longer updated.
• Create a new PAT. Generate a fresh Personal Access Token in TestDino and use it in your MCP config. Tokens from earlier versions will not work with this release.
• New service address. The MCP now talks to mcp.testdino.com. If you set a custom API URL in your config, update it. If you use the default, there is nothing to do.

Switch your config to @testdino/mcp, swap in your new PAT, and you are set.

TestDino MCP v2.0.0

Meet TestDino MCP v2.0.0

TestDino MCP is no longer just about managing and analysing your tests. It now acts on them. From a failed test you can file a tracked bug in your issue tracker, cut through a big red run by grouping failures into real problems, and dig into test history with far more precision. All from natural language, right inside Cursor, Claude Code, or VS Code.

Turn failing tests into tracked issues, without leaving your editor

Connect your team's issue tracker once, then ask your AI assistant to file a ticket straight from a failed test or run. Supported trackers:

• Jira
• Linear
• Asana
• monday.com

What you can now do:

• Connect a tracker by asking your assistant to link the integration (it hands you a secure authorization link to approve in your browser).
• File an issue from a failure with the test context already attached, so the ticket lands with the right details instead of a blank template.
• Check issue status later by asking for updates on tickets you have already filed, including their current state in the external tool. Ask about several at once.
• See what's connected for a project before you start, so you always know which trackers are live.

The result: the gap between "a test failed" and "a ticket exists for it" is now one sentence.

Triage big failing runs by pattern, not one test at a time

When a run fails with dozens of red tests, you no longer have to read them one by one. Ask your assistant to cluster the failures and it groups tests that share the same underlying error. You immediately see which failures are one real problem versus many, so you can fix the root cause first and stop chasing duplicates.

Sharper filtering and history for test runs and cases

Finding the exact run or test you care about is faster now:

• Filter runs by status, by test case tags, by free-text search, and sort the results, so you can zero in on the runs that matter instead of scrolling.
• Filter for artifacts to find tests that actually have screenshots, videos, or traces available, including artifacts from earlier attempts of a flaky test.
• Pull test history and step detail for a case, so you can see how a test has behaved over time and what happened at each step.

Know your access at a glance

The health check now shows your role in the organization, so you can confirm what you are able to do before you try it.

New tools in this release

Five new capabilities your AI assistant can now call:

• connect_integration — link an issue tracker (Jira, Linear, Asana, monday.com) to a project.
• get_integration_status — check which trackers are connected before you act.
• create_external_issue — file a ticket from a failing test or run.
• get_external_issue — pull the current status of tickets you have filed.
• get_run_error_clusters — group a run's failures by their underlying error to triage at scale.

Notes for existing users

This is a major version because a few things changed under the hood:

• Create a new PAT. After updating to 2.0.0, generate a fresh Personal Access Token in TestDino and use it in your MCP config. Tokens from earlier versions will not work with this release.
• New service address. The MCP now talks to mcp.testdino.com. If you set a custom API URL in your config, update it. If you use the default, there is nothing to do.

Update to 2.0.0 with your usual MCP client (or reinstall the latest testdino-mcp package), swap in your new PAT, and you are set.

v1.0.11

What's new?

✨ Smarter test audits. Your AI now knows what's actually broken before it starts reviewing.

Ask your AI assistant for a TestDino audit on your Playwright tests and it
stops guessing. It walks in with the top failing tests, the top flaky
tests, and the top slow tests from your real test runs, for the branch
you're on, before it looks at a single line of your code. The audit that
comes out is grounded in what your test suite is actually doing, not the
AI's hunch.

🧭 The audit conversation, made honest

Before, when you said "audit my checkout flow tests", the AI would read
your files and improvise findings. Now:

1. Your AI pulls context from TestDino first. Which tests keep failing on your branch, which ones are flaky, which ones are getting slow, and how many runs have been recorded lately.
2. Then it reviews your code with that context in hand.
3. Then it files the audit report back to TestDino. Same as before, viewable in the dashboard alongside your test runs.

You'll notice the reports read more like "the checkout total-calculation
test has failed 8 out of the last 10 runs, here's why the assertion is
too loose" instead of generic "consider stronger assertions" advice.

🧩 Two focused tools now drive the audit flow

The audit runs on two tools:

• get_audit_report: fetches the audit context (prompt + branch signals) at the start of a new audit, or browses and retrieves past audit reports.
• submit_audit_report: hands in the completed audit with score, findings, recommendations, and markdown.

Regular conversational usage ("run a TestDino audit on my Playwright
tests") just works, no config changes needed for your Cursor or Claude
Desktop setup.

Heads up for custom automations and CI scripts: if you were calling
test_audit by name in your own scripts, update those references to
get_audit_report and submit_audit_report. The test_audit tool has
been replaced by these two.

Full Changelog

v1.0.10...v1.0.11

v1.0.10

What's new?

✨ 14 new tools — your agent goes from observer to operator

Your AI assistant (Claude, Cursor, VS Code, Claude Desktop, etc.) can now plan releases,
manual tests, and exploratory sessions — all from chat. 27 total tools, up
from 13.

---

🚀 Releases

list_releases, get_release, create_release, update_release

• Browse, view, create, and update releases — set dates, track completion
• Nested releases up to three levels deep (program → release → sub-release)
• Roll-up stats across every run inside a release and its children

"Create a release called Q3 GA starting next Monday, then nest Q3 Hardening under it."

---

🧪 Manual testing — plan, execute, report

End-to-end manual testing over MCP.

• Plan (existing): create_manual_test_case, create_manual_test_suite — draft cases
  with steps, expected results, tags, Jira links; group into suites
• Run (new): create_manual_run, update_manual_run, list_manual_runs,
  get_manual_run, list_run_test_cases, update_run_test_case
• Assign cases and set results like clicking Assign to and tapping the result pill —
  works on untested cases too
• Run protection — closed runs are read-only but can still be moved between releases

"Draft 8 cases for the checkout flow, group them into a Checkout v2 suite, kick off a run
on staging, and split assignments between me and alex@acme.com."

---

🔍 Exploratory sessions

create_session, update_session, list_sessions, get_session

• Rich-text mission, time estimate, type, environment, tags
• Assign members — Add members to contribute in exploratory sessions.
• Findings still live in the UI for v1

"Start a 60-minute session on the new checkout flow, assign it to alex@acme.com, link it to
Q3 GA."

---

🔌 Client compatibility

• Added support for Cursor, VS Code, JetBrains, and Claude Desktop auth-redirect patterns

---

Full Changelog:
v1.0.9...v1.0.10

v1.0.9

What's new?

✨ New Feature: test_audit tool ─

Guide your AI agent through structured Playwright test quality audits via TestDino. The tool runs a server-curated audit flow that fetches a
TestDino-defined prompt + branch signals, lets the agent analyze the local Playwright code, and submits the completed report back to TestDino.

What it does:

• action: "analyze" — Fetches the audit prompt and branch context, then submits the final structured report (score, findings,
  recommendations, markdown) once the agent has produced it
• action: "list" — Browse previously submitted audit reports, optionally filtered by branch
• action: "get" — Retrieve the full contents of a specific past report

Highlights:

• Strict trigger semantics — only fires when the user explicitly names TestDino and the target is Playwright code (detected via
  @playwright/test imports, Playwright APIs, .spec.ts / .spec.js files, or playwright.config.*)
• Optional branch parameter for branch-scoped audit context
• Writes the final markdown report to a local file (default TEST-AUDIT.md) before submission
• Supports both feature-level and spec-file-level scopes

📚 Documentation

• New test_audit section in docs/TOOLS.md
• Expanded docs/skill.md with audit trigger scenarios, Playwright gate rules, and mandatory agent behaviors
• Added .env.example for local dev

🧪 Tests

• 408 lines of unit coverage for test-audit.ts
• Integration test updates for the new tool registration

🛠 Internal

• New endpoint builders for audit context / submit / list / get in endpoints.ts
• Improved request.ts error handling

Full Changelog: v1.0.8...v1.0.9

v1.0.8

What's New

Claude Code Setup

• Added CLAUDE.md with project conventions, tool patterns, and self-learning protocols
• 7 skills (/add-tool, /verify, /review, /publish, /debug-tool, /update-docs, /retro), 2 commands (/bump-version, /investigate-tool), 3 automated hooks
• Issue tracking system (ISSUES.md) with 13 findings from code review

Test Suite

• Switched from Jest to Vitest with v8 coverage
• 92 tests across 17 files (unit + integration)
• Coverage: 90.5% statements, 81.3% branches, 100% functions
• Integration tests verify full MCP server via JSON-RPC protocol

CI/CD

• CI workflow: format, typecheck, lint, test, build across Node 20/22/24
• Publish workflow: triggers on GitHub release, validates version consistency, publishes to npm with provenance

Code Cleanup

• Deduplicated ~130 lines of shared code between manual test case tools
• Removed unused docker dependency and logger.ts
• Fixed lint error, LICENSE copyright, server.json env var name
• Minimum Node.js bumped from 18 to 20 (Node 18 EOL)

Bug Fixes

• Fixed unnecessary async on ReadResourceRequestSchema handler
• Fixed prettier formatting across all source files