Kilo-Org/kilocode v5.9.0

Release v5.9.0

on GitHub

5.9.0

Minor Changes

• #5560 84ab615 Thanks @marciepeters! - Add Poe as a supported API provider

Patch Changes

• #5860 baeb3f4 Thanks @Neonsy! - Fix OpenAI Responses Azure URL normalization so Azure v1 endpoints avoid unsupported api-version parameters.

• #5968 053775e Thanks @Olusammytee! - Normalize legacy Claude Code model IDs so dated aliases resolve to canonical model metadata and preserve capabilities (including image support).

• #5995 a2c7c49 Thanks @mujtaba93! - Fix: Prevent terminal focus stealing in Agent Manager (fixes #5946)

• #5990 7d23f2c Thanks @hdcodedev! - Fix OpenAI-compatible Responses fallback requests when custom base URLs already include /v1 (#5979).

• #5916 8cceb67 Thanks @Githubguy132010! - Fix blank messages and UI not updating when canceling a task in Agent Manager

• #5634 be40801 Thanks @Patel230! - Fix context condensing prompt not saving properly

• #5864 c92c6b1 Thanks @Githubguy132010! - Fixed organization selector overlapping with "Recent" text in chat pane header

• #5267 1467783 Thanks @maywzh! - fix: preserve extra_content for Gemini 3 thought_signature support

• #5569 30eb061 Thanks @romeoscript! - Retry Amazon Bedrock network connection lost errors up till 3 times

• #5992 23a083a Thanks @shssoichiro! - fix: allow dropdowns in Modes modal to be changed

• #5648 af395f8 Thanks @DDU1222! - Feature: add new provider AIHubmix

• #5993 bedf59d Thanks @saneroen! - Use OpenAI Codex OAuth credentials in Agent Manager so ChatGPT Plus/Pro works in agent mode

• #5942 d266500 Thanks @Githubguy132010! - Fixed resumed agent runtime orchestrator tasks so previous task history is preserved.

• #5893 e29306d Thanks @evanjacobson! - Fix duplicate text output when using OpenAI-compatible providers with streaming disabled.

• #5969 902a33f Thanks @Olusammytee! - Normalize Vertex Claude Opus 4.6 legacy aliases to the canonical model ID to prevent invalid API calls and keep UI/runtime model capabilities consistent.

• #5568 82ba1a8 Thanks @romeoscript! - fix: override context window for MiniMax/Kimi free models

• #5885 c7d5865 Thanks @Olusammytee! - fix: cap qwen3-max-thinking max_tokens to provider limit