# What's new

Canonical page: https://www.hopeagent.ai/en/changelog

Current stable release: v0.60.0. Released highlights are listed below; see GitHub for the full history.

## 0.60.0

- Weather queries accept Chinese city names; a missing country name for one location no longer fails the whole batch.
- The README defaults to English, keeps a complete Chinese entry point, and updates chat, design, knowledge and dashboard screenshots.

## 0.59.0

- Adds GPT-6.1 Sol, GPT-6 Sol/Luna and Claude Opus/Sonnet 5.5 presets; removes the retired Together AI Kimi K2.6 Serverless preset.
- After stopping a reply, OpenAI Chat-compatible history no longer includes empty assistant messages, so existing sessions can continue.

## 0.58.0

- When sharing is allowed, matching Global memory can still return when project or Agent memory fills the recall limit.

## 0.57.0

- Expands the sidebar new-project click target and adds space from the scrollbar.

## 0.56.0

- Raised the frontend build memory allowance for the separate Intel Mac installer to reduce build failures; this release adds no app features.
- The Intel x64 DMG is uploaded only after its separate build completes; check the v0.56.0 Release assets before downloading.

## 0.55.0

- Import local Codex conversations from the session sidebar; reimports update source records, which remain searchable, viewable, exportable, and read-only.
- Added a Requesty provider template with managed routing policies, custom model IDs, and a European endpoint option.

## 0.54.0

- Design Space chats use and remember the selected model; child tasks without a dedicated model inherit the parent session model.
- Long tool chains can compact earlier completed calls within the same user turn while retaining the original request and latest complete results.

## 0.53.0

- Fixed valid OpenAI Responses tool calls failing when omitted optional arguments were automatically filled in.

## 0.52.0

- Type `@` to reference an existing regular conversation, allowing Hope Agent to read its context or send it work when asked.
- Updated the image-capable DeepSeek preset, filtered reasoning when Think is off, and improved MCP / ACP connection reliability.

## 0.51.0

- Fixed the build and upload failures affecting native Intel Mac x64 installers.
- Intel installers are uploaded separately after the main release; check for an x64 DMG in the Release before downloading. Apple Silicon, Windows, and Linux download methods are unchanged.

## 0.50.0

- Modified-file labels in replies and workspace files with change snapshots open the saved before-and-after diff; newly created files show all additions.
- Missing historical diffs are clearly indicated; the file menu still offers a preview of the current contents.

## 0.49.0

- External knowledge spaces can be explicitly mounted read/write after enabling writes, while preserving read-only limits and project-inherited permissions.
- Fixed black spaces left by closing macOS fullscreen windows, first-time authorization for URL-only remote MCP servers, and pre-release placeholder text in update notes.

## 0.48.0

- Web fetching checks DNS and SSRF boundaries before every redirect, and search diagnostics omit queries, request URLs, and upstream response bodies.
- The build baseline moves to Node.js 24, with improved external-memory version compatibility and evaluation evidence counting.

## 0.47.0

- Google and OpenRouter add Gemini 3.8 Flash presets, reasoning-level mapping, and cost estimates.

## 0.46.0

- Added GPT-6 Astra and Claude Fable 5.1 presets, with better default-model selection, retry delays, and thinking-history preservation.

## 0.45.0

- The built-in skill installer previews fixed content from GitHub subfolders or local directories; questions can select their waiting policy per call.

## 0.44.0

- Custom models can look up capabilities across providers by ID; catalog prices are applied only after confirmation.

## 0.42.0

- Direct DeepSeek connections add a V4 Flash Vision Exp image-understanding preset.

## 0.39.0

- Cross-chat messages support two-way navigation and side-chat restoration; Ollama is installed manually from its official site, and credential storage is more reliable.

## 0.38.0

- Added /side chats, allowing multiple independent follow-ups to generate alongside the main conversation.

## 0.35.0

- Web reading adds stable snapshots, cursor continuation, dynamic pages, and PDFs, alongside toolchain diagnostics, workspace-specific Hook authorization, and design preview regression checks.

## 0.34.0

- Right-side panels become reorderable tabs, improving file browsing, previews, diff navigation, and narrow-window layouts.

## 0.33.0

- Schedules can run in new or existing chats with full run history. Projects support multiple source folders, with text quoting and preview-based desktop pet imports.

## 0.30.0

- Released native Slack/Telegram streaming, /fork chat branching, remote server updates, and the Nori desktop pet. Unpublished 0.29.0 changes were included in this release.

## Documentation and source

- [Hope Agent — Open-source AI assistant with memory and tools](https://www.hopeagent.ai/en)

- [Install Hope Agent](https://www.hopeagent.ai/en/install)

- [Hope Agent User Guide](https://www.hopeagent.ai/en/guide)

- [Running modes](https://www.hopeagent.ai/en/running-modes)

- [Features](https://www.hopeagent.ai/en/features)

- [Self-Hosting](https://www.hopeagent.ai/en/self-host)

- [Frequently asked questions](https://www.hopeagent.ai/en/faq)

- [GitHub](https://github.com/shiwenwen/hope-agent)

- [Releases](https://github.com/shiwenwen/hope-agent/releases)

- [MIT](https://github.com/shiwenwen/hope-agent/blob/main/LICENSE)