kiro v1.0.182: global hooks, searchable session history, and automatic agent recovery
Amazon's spec-driven IDE shipped v1.0.182 on July 20 with three changes that address friction points in daily use: hooks that apply across all projects without per-repo setup, a panel that makes past sessions searchable, and automatic recovery when the agent's response stream is interrupted. MCP tools also refresh dynamically instead of requiring a restart.
Kiro v1.0.182, released July 20, addresses three friction points that the spec-driven IDE has had since its GA launch: hooks that required per-project setup, sessions that were effectively unsearchable once they scrolled off screen, and agent runs that failed silently when a response stream was interrupted. None of these are headline features, but together they make the IDE noticeably less annoying to use day-to-day.
global hooks
Previously, Kiro hooks — automations triggered by file events, tool interceptions, or prompt submissions — lived in .kiro/hooks/ inside a specific project. That meant any hook you wanted everywhere (a linter, a commit-message formatter, a context injector) had to be duplicated into every repo.
v1.0.182 adds user-level global hooks stored outside any project directory. A global hook applies to all Kiro sessions regardless of what repo is open. Hooks still use the same structured v1 JSON format with natural-language trigger descriptions; the only difference is where the file lives. Per-project hooks continue to work and take precedence when there is a conflict.
searchable session history panel
The session history panel is now searchable. Prior to this release, finding a specific conversation from last week meant scrolling through a chronological list. The new panel lets you filter by keyword across session titles and (where indexed) content. You can also open any session as a full-width editor tab for more room to read through long conversations — accessible via right-click on a chat tab and selecting “Open in Editor.”
For anyone running Kiro on long research or refactoring projects where past sessions carry important context, this is the change that makes the history actually useful rather than a linear scroll target.
automatic recovery from interrupted responses
When a Kiro agent response stream was interrupted — network hiccup, VPN reconnect, corporate proxy timeout — the session previously required a manual restart. v1.0.182 adds automatic recovery: the agent detects the interruption and resumes or retries without user intervention. This pairs with proxy reliability improvements that specifically target corporate proxy environments, where connection drops during long agent runs were a reported pain point.
dynamic MCP tool refresh
MCP tools registered in a session now refresh dynamically when an MCP server updates its tool list. Before this release, Kiro read the available tools at session start and did not pick up changes made to the server while the session was running. If you updated a tool definition or added a new tool, you had to restart the session. The dynamic refresh eliminates that restart requirement for running sessions.
where kiro stands now
Kiro's positioning — spec-first development where the IDE generates a spec document before writing any code — has not changed. But the v1.0.x series has been working through the UX debt that accumulated during the preview period: proxy failures, session continuity, the per-project hook tax. v1.0.182 closes a few of those reliably.
For teams evaluating spec-driven workflows as a structured alternative to open-ended vibe coding, the session history search and global hooks are the two changes most likely to reduce daily friction. The IDE review on this site covers the broader approach.