Assistants with QA context and boundaries
The MCP catalog brings together 65 tools for QA workflows, each subject to user scope and permissions. Assistants can explore accessible public websites and record requirements, stories, test cases, and observed manual results; compatible automated tests run through the Treseko worker.
Story and case generation preserves clarifications and acceptance criteria, improves recovery from incomplete responses, and keeps proposals reviewable before they are applied.
Runs and reports that keep their context
Run saving can recover progress and prevent duplicates. History improves pagination, filtering, and evidence lookup for API and conversational tests. Suites protect changes and respect active builds when archiving cases.
Bug Tracker, Incident Center, inventory, and reports also receive updates to filters, drafts, permissions, links, metrics, and exports. Reports distinguish work still in progress from final results.
An AI engine that requires evidence
Task scheduling handles temporary failures and duplicate submissions more reliably, and preserves tasks when a build is protected. During evaluation, Treseko requires evidence before asserting certain states and checks the goal and effect after recovering an action.
Confidence scores are labeled internal and uncalibrated. They should not be read as probabilities or as a guarantee that a test passed.
Self-hosted updates
The local atomic updater includes volume backups and rollback, and coordinates service health before activating a release. Premium 1.0.6 installations must first install the 1.0.7 bridge according to the documented upgrade path. Installation ping remains off by default, and Bash and PowerShell installers remain available while graphical installers are temporarily withdrawn.