v1.7.0

Fixed

• --dry-run printed the last four characters of your API token. It now shows only the scheme (Authorization: Basic ****)
• keen which answered plain-English questions with unrelated commands, sometimes destructive ones ("close a sprint" suggested deleting a Bitbucket app). It now ranks exactly like keen-mcp's keen_search, includes keen's own commands (transition, find-issues, sprints, stale), and says "no match" instead of guessing
• keen which printed whole API descriptions, several paragraphs each. It now shows one line per command
• A dry run of a list command went on to print "0 results (live)", a skipped-items warning and "source": "live". It now shows only the request; JSON output carries "dry_run": true
• keen analytics --group-by reported every issue as <nil> because Jira keeps status, assignee and type under fields. Dotted paths work (fields.status.name), and for issues so do status, assignee, type, priority and project
• Help for analytics, sync, search and stale showed example words from another product (messages, author_id, channels, transactions). The examples now use Jira's types and run as written

Added

• keen project delete <key> deletes a project (--enable-undo puts it in the recycle bin). It was the one Atlassian operation with no command; keen project delete project-asynchronously <key> still works

Changed

• keen-mcp answers a read repeated within 5 minutes from keen's response cache, as the CLI already did, so AI agents stay under Atlassian's rate limits. Any successful write clears the cache. KEEN_MCP_CACHE=off makes every read live
• The response cache is readable only by your user (it holds Atlassian data)