@qase/mcp-server

Create or update a test case. If `id` is provided, updates the existing case; if omitted, creates a new one. Enum fields (priority, severity, type, etc.) accept both labels ("high", "blocker") and numeric IDs — the server normalizes automatically. If the project has "Test case review" enabled, changes may need to go through a review instead of being written directly — run qase_discover_tools with "review" for those tools.

Direct Qase REST API call for endpoints not covered by other tools. Pass the HTTP method, path (starting with /v1/), and optional body/query. See https://developers.qase.io for API reference. Use this as an escape hatch when the dedicated tools do not cover your use case. Sends JSON only: it cannot upload files, which need multipart/form-data — use qase_attachment_upload for that.

Upload a file attachment and get back its hash, which the `attachments` field of qase_case_upsert, qase_result_record, qase_ci_report, qase_triage_defect, and qase_shared_step_upsert accepts. This is the only way to obtain such a hash — the upload endpoint needs multipart/form-data, which qase_api cannot send. Pass the file as base64 (`file_base64`), which is the only option when the server runs remotely, such as the hosted connector; `file_path` works only when the server runs on the same machine as the file.

Report CI/CD test results in one call: creates a run, records all results, and optionally completes the run. Replaces the 3-4 step manual workflow of create_run → bulk_create_results → complete_run. Designed for CI pipeline integration.

Create or update a defect. If `id` is provided, updates (including status changes and resolve). If omitted, creates a new defect. Set `status: "resolved"` to resolve an existing defect.

Set up a regression test run in one call. Accepts case selection by suite IDs, explicit case IDs, or plan ID. Creates the run and adds all matching cases. Replaces the multi-step workflow of find cases → create run → add cases.

Search for and activate additional Qase tools. By default, only core tools are visible. Use this to find tools for specific needs: deletions, milestone management, attachments, etc. Found tools are automatically activated and become available for use.

Record one or more test results into a run. Accepts an array of results — a single entry uses the single-result API, multiple entries use bulk. Each result must include a status; case_id is recommended.

Get any Qase entity by type and ID. Supports field projection via the `fields` parameter. For project-scoped entities (case, suite, run, result, plan, defect, milestone, environment, shared_step, shared_parameter, configuration), `code` is required. For global entities (user, author, attachment, custom_field), `code` can be omitted.

Get full project context in one call: project details, suites tree, milestones, environments, custom fields, and users. Cached for 5 minutes. Use this as the first call when starting work with a project — it seeds all the metadata the LLM needs without making 6 separate list calls. Each collection returns its first 100 entities by default; the `coverage` field reports `{ total, loaded, truncated }` per collection, so check it before assuming a list is complete. Pass `full: true` to page through everything.

Create or update a test run. If `id` is provided, updates; if omitted, creates.

Create a defect from a test failure. Requires title, actual_result, and severity. Note: the API offers no way to attach existing runs or results to a defect — the runs/results seen on a defect are populated by the test runner when a result is reported as a defect. Reference failing results in actual_result instead.

Get one section of the Qase Query Language (QQL) reference. `topic` is required — ask for the section you need rather than the whole reference. Start with "entities" when you are unsure which fields an entity has (they differ per entity), "aggregation" to count or summarise without paging, and "enumValues" for valid field values.

Search entities using Qase Query Language (QQL) with powerful filtering and cross-project queries