MoonshotAI · liruifengv · May 29, 2026 · May 29, 2026 · chatgpt-codex-connector · May 29, 2026
diff --git a/.changeset/bg-agent-terminal-status.md b/.changeset/bg-agent-terminal-status.md
diff --git a/.changeset/compact-overflow-retry-budget.md b/.changeset/compact-overflow-retry-budget.md
diff --git a/.changeset/constructable-agent-model-provider.md b/.changeset/constructable-agent-model-provider.md
diff --git a/.changeset/context-overflow-responses.md b/.changeset/context-overflow-responses.md
diff --git a/.changeset/env-var-model-config.md b/.changeset/env-var-model-config.md
diff --git a/.changeset/experimental-flags.md b/.changeset/experimental-flags.md
diff --git a/.changeset/fix-native-update-false-success.md b/.changeset/fix-native-update-false-success.md
diff --git a/.changeset/fix-resume-missing-session-footer-leak.md b/.changeset/fix-resume-missing-session-footer-leak.md
diff --git a/.changeset/fix-ripgrep-cross-device-copy.md b/.changeset/fix-ripgrep-cross-device-copy.md
diff --git a/.changeset/footer-rotating-tips.md b/.changeset/footer-rotating-tips.md
diff --git a/.changeset/plan-usage-percent-used.md b/.changeset/plan-usage-percent-used.md
diff --git a/.changeset/plugin-install-from-github.md b/.changeset/plugin-install-from-github.md
diff --git a/.changeset/preserve-unsigned-thinking.md b/.changeset/preserve-unsigned-thinking.md
diff --git a/.changeset/project-context-messages.md b/.changeset/project-context-messages.md
diff --git a/.changeset/restrict-plugin-trust-badges.md b/.changeset/restrict-plugin-trust-badges.md
diff --git a/.changeset/resume-cross-workdir-sessions.md b/.changeset/resume-cross-workdir-sessions.md
diff --git a/.changeset/retry-model-token-limit.md b/.changeset/retry-model-token-limit.md
diff --git a/.changeset/retry-terminated-stream.md b/.changeset/retry-terminated-stream.md
diff --git a/.changeset/slim-llm-request-log.md b/.changeset/slim-llm-request-log.md
diff --git a/.changeset/split-runtime-kaos.md b/.changeset/split-runtime-kaos.md
diff --git a/.changeset/subagent-stop-wording.md b/.changeset/subagent-stop-wording.md
diff --git a/.changeset/tool-support-services.md b/.changeset/tool-support-services.md
diff --git a/.changeset/unlimited-max-steps-per-turn.md b/.changeset/unlimited-max-steps-per-turn.md
diff --git a/.changeset/use-generic-datasource-stock.md b/.changeset/use-generic-datasource-stock.md
diff --git a/.changeset/user-prompt-hook-context.md b/.changeset/user-prompt-hook-context.md
diff --git a/apps/kimi-code/CHANGELOG.md b/apps/kimi-code/CHANGELOG.md
@@ -1,5 +1,57 @@
 # @moonshot-ai/kimi-code
 
+## 0.6.0
+
+### Minor Changes
+
+- [#212](https://github.com/MoonshotAI/kimi-code/pull/212) [`2bbea75`](https://github.com/MoonshotAI/kimi-code/commit/2bbea75ee4c0b11f12d2921061774426df40479a) - Add a `KIMI_MODEL_*` environment-variable channel that lets you run Kimi Code against a specific model (provider type, base URL, API key, context size, capabilities, and thinking settings) without editing `config.toml`.
+
+- [#221](https://github.com/MoonshotAI/kimi-code/pull/221) [`bab2da7`](https://github.com/MoonshotAI/kimi-code/commit/bab2da7b1c785d6deba25decb1411f8f5a70de8c) - Install plugins directly from GitHub repository URLs, and surface each install's origin and trust level (kimi-official, curated, third-party) in the plugin manager.
+
+- [#118](https://github.com/MoonshotAI/kimi-code/pull/118) [`8913440`](https://github.com/MoonshotAI/kimi-code/commit/891344054111a05171963cfa524ef749c2855321) - Support querying sessions by sessionId or workDir in listSessions, and show a helpful cd command when resuming a session from a different working directory.
+
+- [#186](https://github.com/MoonshotAI/kimi-code/pull/186) [`537cf20`](https://github.com/MoonshotAI/kimi-code/commit/537cf20d18b26d4238f963f793f8a8ef085ac97e) - Remove the default per-turn step limit of 1000. Users can still set `max_steps_per_turn` in config to enforce a custom limit.
+
+### Patch Changes
+
+- [#197](https://github.com/MoonshotAI/kimi-code/pull/197) [`f3269ea`](https://github.com/MoonshotAI/kimi-code/commit/f3269eacb9da9a6b66f578a864d0b9bdfb1d6d81) - Show the real terminal status of background agents in the transcript so lost, failed, and killed ones no longer appear as completed, and include the resume agent id and recovery instructions in the failure notification so the model can resume reliably.
+
+- [#211](https://github.com/MoonshotAI/kimi-code/pull/211) [`54590d3`](https://github.com/MoonshotAI/kimi-code/commit/54590d3d464b05eed0837a725b37f3aa491c09af) - Back off failed compaction retries by a fixed slice of the model context window.
+
+- [#167](https://github.com/MoonshotAI/kimi-code/pull/167) [`b5981a5`](https://github.com/MoonshotAI/kimi-code/commit/b5981a523b66ff2fd5f09a7e66075628b94683c8) - Introduce `ModelProvider` interface and `SingleModelProvider` to decouple `Agent` from `ProviderManager`.
+
+- [#213](https://github.com/MoonshotAI/kimi-code/pull/213) [`2388f20`](https://github.com/MoonshotAI/kimi-code/commit/2388f20bb3d039e89caefca159801059b90dc64a) - Handle context overflow errors consistently across provider responses.
+
+- [#214](https://github.com/MoonshotAI/kimi-code/pull/214) [`caaa6d8`](https://github.com/MoonshotAI/kimi-code/commit/caaa6d83ee262ba4c954386458ee13aacdb26e1a) - Fix the native self-updater reporting a successful update when the install command actually failed.
+
+- [#202](https://github.com/MoonshotAI/kimi-code/pull/202) [`14a0348`](https://github.com/MoonshotAI/kimi-code/commit/14a03488555682dde4bcd74aadf79f60a9827304) - Fix footer leaking onto the terminal when resuming a non-existent session.
+
+- [#198](https://github.com/MoonshotAI/kimi-code/pull/198) [`8c77cfa`](https://github.com/MoonshotAI/kimi-code/commit/8c77cfab62617e07b38f8514a8ef7cddfd9f1069) - Fix automatic ripgrep installation when temporary files are on another filesystem.
+
+- [#199](https://github.com/MoonshotAI/kimi-code/pull/199) [`588145d`](https://github.com/MoonshotAI/kimi-code/commit/588145dc9b266456bdb1d739975a5b9cf33d70ae) - Expand the footer's rotating tips to surface more commands and shortcuts, featuring newer and important ones more prominently.
+
+- [#192](https://github.com/MoonshotAI/kimi-code/pull/192) [`64964a0`](https://github.com/MoonshotAI/kimi-code/commit/64964a0dda98fc2db5e15ba923ea9414c78e0009) - Improve the usage information display in the TUI.
+
+- [#195](https://github.com/MoonshotAI/kimi-code/pull/195) [`3a0e060`](https://github.com/MoonshotAI/kimi-code/commit/3a0e06031ac6dfde148f64906a06cfe820ad9c63) - Project persisted hook and blocked prompt messages into model context.
+
+- [#221](https://github.com/MoonshotAI/kimi-code/pull/221) [`bab2da7`](https://github.com/MoonshotAI/kimi-code/commit/bab2da7b1c785d6deba25decb1411f8f5a70de8c) - Restrict plugin trust badges to Kimi-hosted plugin CDN URL patterns.
+
+- [#207](https://github.com/MoonshotAI/kimi-code/pull/207) [`e280f33`](https://github.com/MoonshotAI/kimi-code/commit/e280f33daf7fbf1271c872dcb224737ec9518f73) - Recover from provider model token limit errors during long conversations.
+
+- [#201](https://github.com/MoonshotAI/kimi-code/pull/201) [`3da4dae`](https://github.com/MoonshotAI/kimi-code/commit/3da4daeadee39573c7eeede30fa9465b411be3e2) - Automatically retry when a model response stream is dropped mid-flight (a `terminated` error) instead of failing the turn.
+
+- [#190](https://github.com/MoonshotAI/kimi-code/pull/190) [`1873859`](https://github.com/MoonshotAI/kimi-code/commit/1873859b0ef093a956dfd19e1530e920e7118160) - Slim the LLM diagnostic logs with fewer, more compact fields.
+
+- [#185](https://github.com/MoonshotAI/kimi-code/pull/185) [`114777e`](https://github.com/MoonshotAI/kimi-code/commit/114777e859680f807375760271533e2dc396af5d) - Split `RuntimeConfig` into `Kaos` and `ToolServices` and update all references accordingly.
+
+- [#189](https://github.com/MoonshotAI/kimi-code/pull/189) [`564721f`](https://github.com/MoonshotAI/kimi-code/commit/564721fe16e582b2774835b01dec799cbb1d0122) - Clarify subagent and background task stop messages as user-initiated.
+
+- [#206](https://github.com/MoonshotAI/kimi-code/pull/206) [`07d51e4`](https://github.com/MoonshotAI/kimi-code/commit/07d51e4add6ee23a56fb8745aa7754f05f3d6d36) - Relocate shared tool service typing to the tool support layer.
+
+- [#215](https://github.com/MoonshotAI/kimi-code/pull/215) [`b9860e9`](https://github.com/MoonshotAI/kimi-code/commit/b9860e9f6ec65eb5dfdabbad54f1a87d69f4f00a) - Align the datasource plugin with the generic two-tool workflow.
+
+- [#200](https://github.com/MoonshotAI/kimi-code/pull/200) [`5159af3`](https://github.com/MoonshotAI/kimi-code/commit/5159af341c7d388a158e41afb470a2281333f329) - Keep blocked prompt hook conversations available to subsequent model turns.
+
 ## 0.5.0
 
 ### Minor Changes

diff --git a/apps/kimi-code/package.json b/apps/kimi-code/package.json
@@ -1,6 +1,6 @@
 {
   "name": "@moonshot-ai/kimi-code",
-  "version": "0.5.0",
+  "version": "0.6.0",
   "description": "The Starting Point for Next-Gen Agents",
   "license": "MIT",
   "author": "Moonshot AI",

diff --git a/flake.nix b/flake.nix
@@ -190,7 +190,7 @@
               inherit (finalAttrs) pname version src pnpmWorkspaces;
               inherit pnpm;
               fetcherVersion = 3;
-              hash = "sha256-dht627zodv/xeefY7fEvSaKXavTxi+JWA3gwL18U8dk=";
+              hash = "sha256-HpRlxlXZoVqAzrdMdSWhLcTRM1DvDvytVbzIGBo8QUo=";
             };
 
             nativeBuildInputs = [

diff --git a/packages/agent-core/CHANGELOG.md b/packages/agent-core/CHANGELOG.md
@@ -1,5 +1,50 @@
 # @moonshot-ai/agent-core
 
+## 0.5.0
+
+### Minor Changes
+
+- [#212](https://github.com/MoonshotAI/kimi-code/pull/212) [`2bbea75`](https://github.com/MoonshotAI/kimi-code/commit/2bbea75ee4c0b11f12d2921061774426df40479a) - Add a `KIMI_MODEL_*` environment-variable channel that lets you run Kimi Code against a specific model (provider type, base URL, API key, context size, capabilities, and thinking settings) without editing `config.toml`.
+
+- [#205](https://github.com/MoonshotAI/kimi-code/pull/205) [`96bbc47`](https://github.com/MoonshotAI/kimi-code/commit/96bbc471c4aca9526e4dcfe00e6bad2b653bbe66) - Add an experimental feature-flag system: a central registry (`flags/registry.ts`) plus an env-driven resolver. Gate a feature with `flags.enabled('id')`, toggled via `KIMI_CODE_EXPERIMENTAL_<NAME>` or the `KIMI_CODE_EXPERIMENTAL_FLAG` master switch. No flags are defined yet.
+
+- [#221](https://github.com/MoonshotAI/kimi-code/pull/221) [`bab2da7`](https://github.com/MoonshotAI/kimi-code/commit/bab2da7b1c785d6deba25decb1411f8f5a70de8c) - Install plugins directly from GitHub repository URLs, and surface each install's origin and trust level (kimi-official, curated, third-party) in the plugin manager.
+
+- [#118](https://github.com/MoonshotAI/kimi-code/pull/118) [`8913440`](https://github.com/MoonshotAI/kimi-code/commit/891344054111a05171963cfa524ef749c2855321) - Support querying sessions by sessionId or workDir in listSessions, and show a helpful cd command when resuming a session from a different working directory.
+
+- [#186](https://github.com/MoonshotAI/kimi-code/pull/186) [`537cf20`](https://github.com/MoonshotAI/kimi-code/commit/537cf20d18b26d4238f963f793f8a8ef085ac97e) - Remove the default per-turn step limit of 1000. Users can still set `max_steps_per_turn` in config to enforce a custom limit.
+
+### Patch Changes
+
+- [#197](https://github.com/MoonshotAI/kimi-code/pull/197) [`f3269ea`](https://github.com/MoonshotAI/kimi-code/commit/f3269eacb9da9a6b66f578a864d0b9bdfb1d6d81) - Show the real terminal status of background agents in the transcript so lost, failed, and killed ones no longer appear as completed, and include the resume agent id and recovery instructions in the failure notification so the model can resume reliably.
+
+- [#211](https://github.com/MoonshotAI/kimi-code/pull/211) [`54590d3`](https://github.com/MoonshotAI/kimi-code/commit/54590d3d464b05eed0837a725b37f3aa491c09af) - Back off failed compaction retries by a fixed slice of the model context window.
+
+- [#167](https://github.com/MoonshotAI/kimi-code/pull/167) [`b5981a5`](https://github.com/MoonshotAI/kimi-code/commit/b5981a523b66ff2fd5f09a7e66075628b94683c8) - Introduce `ModelProvider` interface and `SingleModelProvider` to decouple `Agent` from `ProviderManager`.
+
+- [#213](https://github.com/MoonshotAI/kimi-code/pull/213) [`2388f20`](https://github.com/MoonshotAI/kimi-code/commit/2388f20bb3d039e89caefca159801059b90dc64a) - Handle context overflow errors consistently across provider responses.
+
+- [#198](https://github.com/MoonshotAI/kimi-code/pull/198) [`8c77cfa`](https://github.com/MoonshotAI/kimi-code/commit/8c77cfab62617e07b38f8514a8ef7cddfd9f1069) - Fix automatic ripgrep installation when temporary files are on another filesystem.
+
+- [#195](https://github.com/MoonshotAI/kimi-code/pull/195) [`3a0e060`](https://github.com/MoonshotAI/kimi-code/commit/3a0e06031ac6dfde148f64906a06cfe820ad9c63) - Project persisted hook and blocked prompt messages into model context.
+
+- [#221](https://github.com/MoonshotAI/kimi-code/pull/221) [`bab2da7`](https://github.com/MoonshotAI/kimi-code/commit/bab2da7b1c785d6deba25decb1411f8f5a70de8c) - Restrict plugin trust badges to Kimi-hosted plugin CDN URL patterns.
+
+- [#207](https://github.com/MoonshotAI/kimi-code/pull/207) [`e280f33`](https://github.com/MoonshotAI/kimi-code/commit/e280f33daf7fbf1271c872dcb224737ec9518f73) - Recover from provider model token limit errors during long conversations.
+
+- [#190](https://github.com/MoonshotAI/kimi-code/pull/190) [`1873859`](https://github.com/MoonshotAI/kimi-code/commit/1873859b0ef093a956dfd19e1530e920e7118160) - Slim the LLM diagnostic logs with fewer, more compact fields.
+
+- [#185](https://github.com/MoonshotAI/kimi-code/pull/185) [`114777e`](https://github.com/MoonshotAI/kimi-code/commit/114777e859680f807375760271533e2dc396af5d) - Split `RuntimeConfig` into `Kaos` and `ToolServices` and update all references accordingly.
+
+- [#189](https://github.com/MoonshotAI/kimi-code/pull/189) [`564721f`](https://github.com/MoonshotAI/kimi-code/commit/564721fe16e582b2774835b01dec799cbb1d0122) - Clarify subagent and background task stop messages as user-initiated.
+
+- [#206](https://github.com/MoonshotAI/kimi-code/pull/206) [`07d51e4`](https://github.com/MoonshotAI/kimi-code/commit/07d51e4add6ee23a56fb8745aa7754f05f3d6d36) - Relocate shared tool service typing to the tool support layer.
+
+- [#200](https://github.com/MoonshotAI/kimi-code/pull/200) [`5159af3`](https://github.com/MoonshotAI/kimi-code/commit/5159af341c7d388a158e41afb470a2281333f329) - Keep blocked prompt hook conversations available to subsequent model turns.
+
+- Updated dependencies [[`2388f20`](https://github.com/MoonshotAI/kimi-code/commit/2388f20bb3d039e89caefca159801059b90dc64a), [`13e0fff`](https://github.com/MoonshotAI/kimi-code/commit/13e0fff462e2ddbec5fb4c9de8ed8e6068db09f1), [`e280f33`](https://github.com/MoonshotAI/kimi-code/commit/e280f33daf7fbf1271c872dcb224737ec9518f73), [`3da4dae`](https://github.com/MoonshotAI/kimi-code/commit/3da4daeadee39573c7eeede30fa9465b411be3e2)]:
+  - @moonshot-ai/kosong@0.2.3
+
 ## 0.4.0
 
 ### Minor Changes

diff --git a/packages/agent-core/package.json b/packages/agent-core/package.json
@@ -1,6 +1,6 @@
 {
   "name": "@moonshot-ai/agent-core",
-  "version": "0.4.0",
+  "version": "0.5.0",
   "private": true,
   "description": "The unified agent engine for Kimi",
   "license": "MIT",

diff --git a/packages/kosong/CHANGELOG.md b/packages/kosong/CHANGELOG.md
@@ -1,5 +1,17 @@
 # @moonshot-ai/kosong
 
+## 0.2.3
+
+### Patch Changes
+
+- [#213](https://github.com/MoonshotAI/kimi-code/pull/213) [`2388f20`](https://github.com/MoonshotAI/kimi-code/commit/2388f20bb3d039e89caefca159801059b90dc64a) - Handle context overflow errors consistently across provider responses.
+
+- [#222](https://github.com/MoonshotAI/kimi-code/pull/222) [`13e0fff`](https://github.com/MoonshotAI/kimi-code/commit/13e0fff462e2ddbec5fb4c9de8ed8e6068db09f1) - Preserve unsigned assistant thinking when serializing history for the Anthropic provider, instead of dropping it. Anthropic-compatible backends (e.g. Kimi) stream thinking without a signature yet reject a tool-call turn whose thinking is missing ("thinking is enabled but reasoning_content is missing"). api.anthropic.com always supplies a signature, so its behavior is unchanged.
+
+- [#207](https://github.com/MoonshotAI/kimi-code/pull/207) [`e280f33`](https://github.com/MoonshotAI/kimi-code/commit/e280f33daf7fbf1271c872dcb224737ec9518f73) - Recover from provider model token limit errors during long conversations.
+
+- [#201](https://github.com/MoonshotAI/kimi-code/pull/201) [`3da4dae`](https://github.com/MoonshotAI/kimi-code/commit/3da4daeadee39573c7eeede30fa9465b411be3e2) - Automatically retry when a model response stream is dropped mid-flight (a `terminated` error) instead of failing the turn.
+
 ## 0.2.2
 
 ### Patch Changes

diff --git a/packages/kosong/package.json b/packages/kosong/package.json
@@ -1,6 +1,6 @@
 {
   "name": "@moonshot-ai/kosong",
-  "version": "0.2.2",
+  "version": "0.2.3",
   "private": true,
   "description": "The LLM abstraction layer for modern AI agent applications",
   "license": "MIT",