v1.0.82026-11-17
Eigent v1.0.8 重做了应用的导航方式:左侧新增一条窄栏,收纳主页、置顶的 Space、设置和账户;主页集中列出所有 Space 及其 Session,可直接打开、新建、置顶、重命名、结束或删除,右键呼出菜单,未读动态一眼可见。按 ⌘K 或点顶栏搜索,可跨 Space 查找 Space、Session、任务、文件、Automations 和 Dispatch;顶栏还加入了前进/后退。
新版会话时间线改为默认:智能体的工作以任务日志呈现,审批与提问嵌在对话里并标明对应的文件或命令,模型切换处有分隔线,详略可调。输入框合并为一个「+」按钮,统一添加文件、Skills 和 Connectors。Space 可用 emoji 代替首字母,选择器支持离线。
修复方面:Space 可删除仍有 Session 的情况、可使用主目录外的文件夹(系统目录仍拒绝);参数畸形的工具调用在执行前被拦截,任务不再中断;浏览器与搜索工具忽略多余参数照常运行;内存不再把 Automations 的消息当作用户发言;Pro 试用额度耗尽时给出提示,刷新不再卡死;中断任务不再反复重载状态。另有文件预览权限收紧、单次写入失败不再拖垮整个工作区、Git 仓库下并发启动 Session 等一批界面与稳定性修补。
v1.0.72026-10-30
Eigent v1.0.7 支持多 Session 并行:同一 Space 内最多并排三个,也能单独开窗口或浏览器标签页;后台的定时任务只从一个窗口启动,不会重复执行。新增 Session 菜单(置顶、重命名、结束、删除,或转成技能与自动化)和消息操作(复制、编辑重发、展开长消息),并加入 Space 概览面板、重新设计的 Dispatch 页面,以及 BytePlus ModelArk 子代理接入。
自动化修复较多:「立即运行」真正即时生效,漏收的运行会补回;按整个 Space 列表展示,时间以本地时区为准;打开对应 Session 后不会重复触发。
审批与提问移到输入框上方并标明涉及的文件或命令,重载后仍可作答;失败任务的后续消息排队等待,恢复的任务保留浏览器工具,中断任务会列出已完成动作与写入文件。退出时若有任务在进行会先询问并清理终端环境;更新后保持登录并恢复上次的 Space;Session 内切换模型只影响当前 Space。界面方面加宽了聊天栏、对齐了消息与输入控件,修复了 PDF 预览、滚动条、图标线宽等细节。
v1.0.62026-10-16
Eigent v1.0.6 重做了自动化与模型设置:自动化页新增六个示例工作流、运行看板与本地时间预览,可暂停、恢复或手动触发,但定时任务需保持应用开启;模型配置统一收进 Settings → Models,支持搜索、手动添加与默认模型选择。引导流程新增「工作角色」步骤,可跳过。任务结果现在会说明完成、停止、失败、超时或中断的原因,失败时还能查看已记录动作。
Session 连续性明显改善:切换会话保留未发送文本与终端回滚,窗口重载后可继续发消息,审批与计时器在重载后也能恢复。模型兼容性方面,未注册思考强度的模型改用提供商默认值,不再被阻断;同时补充了 GPT-6 Sol 的能力声明。文件与空间方面修复了 PDF 空白预览、清理了已删除会话数据,并加强了旧版本地路由的鉴权。
v1.0.52026-09-25
Eigent v1.0.5 improves Session recovery, task queues, and the tools you use to inspect ongoing work. It also adds Meta Model API support, makes Space resources easier to configure, and introduces a limited, opt-in preview of parallel Session execution.
Highlights
- More reliable Session recovery — Recover task state, conversation history, elapsed time, and completed output more consistently after interrupted connections or app restarts. Resume Sessions with their accepted model configuration and handle model switching after usage limits more reliably.
- A clearer task queue — Reorder queued tasks with drag-and-drop or keyboard controls, choose which task starts next, and explicitly stop the current task before starting another. Clearer waiting, stopping, and starting states make the queue easier to follow.
- Background processes and better previews — Inspect running processes, open their terminal output or local server previews, and stop them from the Session side panel. Improve PDF rendering, preview failure recovery, and local browser loading; HTML previews now support files up to 10 MiB.
- Easier Space configuration — Browse configured Skills, Connectors, MCP servers, and models from Space settings. Apply a published and set-up Space's default model to new Sessions, while preserving manual Session selections and existing Session models.
- Expanded model support — Add Meta Model API and Muse Spark to custom model settings. Improve model discovery, API-key errors, reasoning-effort handling, OpenAI/Azure tool requests, Anthropic-compatible tool calls, and OpenAI Workforce initialization with saved credentials.
Reliability and usability improvements
- Show clearer credit, access, connection, and provider errors, with fewer duplicate reminders.
- Preserve completed task results, file summaries, and historical durations more consistently; stop timers for failed historical tasks.
- Keep long-running trigger executions from being marked failed solely because a server-side time threshold elapsed.
- Improve workspace write coordination, process shutdown, provider retries, cancellation, and incremental memory processing.
- Complete Space deletion even when the local backend is temporarily unavailable.
- Improve Skills file browsing, reduce Connector loading flicker, and keep Settings headers visible.
- Separate network proxy Save, Reset, and Restart actions, and prevent edits while a save is pending.
- Connect thumbs-up and thumbs-down feedback to message-level events.
- Handle Git filenames containing spaces, Unicode, or line breaks correctly.
- Bundle Git with Windows x64 builds to prevent backend startup failures when Git is not installed.
- Save credential environment files atomically and enforce owner-only permissions on supported systems.
Parallel Session execution preview
This release includes a default-off, opt-in preview for supported local Single Agent file workflows. On explicitly configured deployments, eligible new Sessions can run concurrently in isolated workspaces, with results automatically integrated back into their Space. Tasks within each Session still run in order.
The preview requires deployment setup, a supported custom or local model, existing file permissions, and explicit opt-in for each new Session. It does not include existing Sessions, attachments, MCP, general terminal/browser tools, Workforce, or cloud execution. See the preview setup and scope.
Upgrade notes
Existing Sessions keep their current execution path; upgrading does not automatically enable parallel execution. Some older interrupted tasks may require a new task if their saved model capabilities no longer match the current configuration.
Full changelog: v1.0.4 → v1.0.5
v1.0.42026-09-04
Eigent v1.0.4
Eigent v1.0.4 introduces a redesigned way to manage Skills and Connectors, while strengthening reliability across multi-turn agent work, workspace checkpoints, and app shutdown.
The new management experience makes resources easier to browse, inspect, and configure, while a focused set of fixes helps Tasks complete more predictably.
Highlights
- Redesigned Skills and Connectors
- Browse and manage Skills in a library-style dashboard with dedicated detail views.
- Inspect source and access details, enable or disable Skills, and review their files from one place.
- Manage Connectors through a clearer collection, discovery, and detail flow.
- A more consistent Home and Settings experience
- Use shared layouts, breadcrumbs, sidebars, and content rails across Home, Skills, and Connectors.
- Navigate clearer lists, empty states, and Space detail views.
- More reliable multi-turn agent work
- Preserve trusted agent instructions across chained Responses API turns.
- Avoid duplicated prompt content while keeping Prompt Guard protections in place.
- Improve compatibility with tools that use dynamic parameter schemas.
- Safer Git-backed Spaces
- Treat nested Git repositories as independent boundaries during workspace checkpoints.
- Complete parent checkpoints without altering nested repositories or their committed content.
- Clearer connector behavior
- Keep Web search discoverable when it still needs configuration.
- Present Slack through the appropriate connector path when Connector Gateway is enabled, while preserving local-only compatibility.
Reliability and UX improvements
- Quit cleanly when the final app window closes and shut down the local backend correctly across platforms.
- Give stopped preview servers more time to release workspace writes and complete checkpointing before Tasks finish.
- Improve generated-web Task completion and background-process cleanup.
- Preserve compatibility for existing saved Workers, Slack triggers, and local-only integrations.
Compatibility
Existing Spaces, Sessions, Tasks, Skill settings, saved Workers, and connector configurations remain available after upgrading. No additional configuration is required for the default v1.0.4 experience.
Full changelog: https://github.com/eigent-ai/eigent/compare/v1.0.3...v1.0.4
v1.0.32026-08-28
Eigent v1.0.3
Eigent v1.0.3 makes long-running work more reliable and turns each Space into a durable, versioned workspace.
Tasks can recover from interruptions, generated files are easier to inspect and review, and reusable workspace configurations help teams start complex work with less setup.
Highlights
- Durable Tasks and recovery
- Resume interrupted Tasks after refreshing or restarting the app.
- Preserve completed work without automatically repeating external actions.
- Keep Task history, elapsed time, approvals, and recovery state available across sessions.
- Git-backed Spaces
- Automatically initialize version history for Eigent-managed Spaces.
- Reuse an existing Git repository when a selected folder already contains one.
- Review Task-scoped file changes, diffs, checkpoints, branches, and version history.
- In-session Browser, Files, and Terminal
- Open web pages, generated files, and an interactive local terminal without leaving the Session.
- Preserve preview tabs and terminal state while switching between Sessions.
- Inspect live and historical terminal output produced by agents.
- Improved file previews
- Preview local HTML, images, PDFs, text files, and larger generated artifacts.
- Open files normally or use View changes for a Task-scoped diff review.
- Workspace Bundles
- Create reusable configurations containing agents, skills, MCPs, model profiles, context sources, permissions, environment requirements, and secret slots.
- Validate and configure Bundles before running them.
- Scoped Memory
- Save reusable notes at User, Space, and Session levels.
- Search, organize, confirm, and archive Memory while keeping canonical Task history separate.
- New model provider
- Added Ant Ling as a built-in OpenAI-compatible provider.
Reliability and UX improvements
- Improved approval, cancellation, timeout, and interrupted-Task handling.
- Added safer recovery when a tool may have produced an external side effect.
- Improved generated-file discovery and Task-scoped change boundaries.
- Hardened terminal lifecycle management and filtered credential-like environment values.
- Fixed the folded Task view scrolling unexpectedly.
- Improved compatibility with existing Sessions and historical Task data.
Compatibility
Existing Spaces, Sessions, and Task history remain available after upgrading. No additional configuration is required for the default v1.0.3 experience.
Full changelog: https://github.com/eigent-ai/eigent/compare/v1.0.2...v1.0.3
v1.0.22026-07-21
Eigent v1.0.2
This release focuses on workflow reliability, connector setup, and desktop app polish.
Highlights
- Added Connector Gateway UI and MCP wiring.
- Added typed workflow event infrastructure for app-level workflow events.
- Improved the in-session preview panel experience.
- Improved active form / single-agent work-log display.
- Added support for BYOK model parameters.
- Added clipboard image paste support in the chat box.
- Improved generated file cards and sandbox path handling.
- Improved update UX and log download flow.
- Refined chat box bottom area interactions and tooltip design.
Reliability Fixes
- Fixed active Space hydration behavior.
- Prevented project switches from killing live runs.
- Allowed remote control background projects to continue working correctly.
- Fixed local sign-in losing the user account state in an empty read-only workspace.
- Ignored stale human replies after task lock cleanup.
- Fixed completed tasks being left as ongoing when answers exceed summary length.
- Fixed user-scoped project output file resolution.
- Hardened local backend startup.
- Fixed backend system trust store handling.
- Fixed Windows terminal capability detection.
- Fixed OAuth authorize URL parameter encoding.
- Kept .git in the build context so server Docker builds succeed.
- Fixed invisible chat code block text under mismatched OS/app theme settings.
- Fixed chat input caret jumps during fast typing.
- Restored the last visited session at launch.
- Stabilized the Single Agent option in the skill scope picker.
Dependencies and Docs
- Bumped camel-ai to 0.2.91a5.
- Updated server README Python requirement from 3.10 to 3.12.
- Updated the WeChat QR code.
v1.0.12026-07-01
v1.0.02026-06-16
v0.0.912026-05-19