OpenHands
106 versions indexed · latest v1.43.0 on Nov 20, 2026
202666
feat(automations): create automations enabled from the setup wizard · feat(settings): Features page for experimental user feature flags, with Conversation reactions · feat(chat): render experimental decision cards
feat(cloud): pick the exact connected git provider for repositories · feat(verify-openhands): every run has a non-loopback plain-HTTP origin
feat(verify-openhands): scenarios beyond the happy path, map controls, browser timezone · feat(acp): choose the reasoning effort from what the agent reports
feat: improve Codex subscription authentication guidance
feat(automations): filter the dashboard by owner with a Service accounts option · feat(chat): move the message hover timestamp closer to its message · feat(conversation): keep Stop hook feedback out of the chat
feat: edit executable plugin configuration in automation help · feat: display conversation budgets with semantic meter fills · feat: add Paper theme to Agent Canvas
feat(automations): explain each automation with a Mermaid diagram
Temporary evidence assets for OpenHands/OpenHands#19394. Browser click-through of automation setup dead ends and decoded PostHog automation_setup_blocked payloads.
feat(automations): open templates without the missing-integration notice · feat(automations): list a script automation's files collapsed
feat(home): remember and clear the cloud repository selection · feat(automations): store picked Bitbucket Data Center repositories as clone URLs · feat(automations): list each event source's own event types
feat(verify): stand in for DigitalOcean so Managed Agents can be verified · feat(canvas): per-conversation Docker runtime and Settings → Runtime page · feat(automations): filter and edit automation tags
feat(verify-openhands): launch --mars-web and --lock-to-cloud · feat(digitalocean): keep the web build signed in across restarts
feat(home): store pinned automations in agent-server settings · feat(mcp): let the install dialog choose between installable library options
feat(automations): hide the GitHub token secret field on self-hosted instances when GitHub is connected (OHE-3699)
refactor(diff): express the Markdown preview cap in CSS · refactor(automations): express the activity track height in CSS · chore: bump agent-server to 1.57.0 · chore: remove merged PR artifacts
feat(automations): reflect the built-in integration's connection state in the setup choice modal (OHE-3612)
feat(backends): show full organization names in the backend selector · feat(launchers): read public skills from the bundled extensions package · feat(agent-profiles): choose skills per profile from the server catalog
chore: bump agent-server to 1.56.0 · chore: remove merged PR artifacts
feat(verify-openhands): executable known bugs and a repeatable fix proof · feat(i18n): add Bahasa Indonesia (id) locale · feat: add close control to Getting started checklist
fix(docker): copy MARS web bridge modules into the image · fix(themes): make warning feedback theme-aware
feat(verify-openhands): drive the mock-LLM e2e behaviors on DeepSeek and map the gaps · feat(verify-openhands): record videos of the browser with browser record
feat(apps): add authenticated backend view helper · feat(chat): render empty-response corrective nudge as an informational note · feat: verify-openhands skill with control-openhands CLI and the full feature map
feat: outline the back navigation control on settings sub-pages · feat(settings): bulk-add provider models as LLM profiles · feat: add direct-prompt Model Router settings
feat: toggle all workspace folders from the Conversations header · feat: scope OpenAI subscription caches by backend · feat(home): reframe splash copy around day-to-day engineering work
feat(desktop): universal macOS DMG with per-arch bundled runtimes · feat: Support Light+ and Solarized Light theme · feat: tag Agent Canvas telemetry by deployment kind (OSS-13530)
feat: test remote MCP servers on cloud backends via the app server · feat: open Git Sync to org admins on cloud backends · feat(i18n): translate recently added Turkish UI strings
feat: request Docker execution workspaces from Docker servers
feat(agent-profiles): select secrets available to a profile · feat: forward Docker conversation runtime settings · feat(automations): select saved agent profiles
feat: scope an agent profile to specific MCP servers · feat: add GPT-6 Astra model support · feat: track authenticated Canvas arrivals · feat: surface automation disablement reason in the UI
feat: click an image attachment to open it full size · feat: only the creator can turn an automation back on · feat: allow editing automations on cloud backends · feat: show which identity an automation runs as
feat: show manifest-declared value statements on dashboard cards and rows · feat: Make automation run UI task-outcome aware · feat: enable LLM provider connections on cloud
feat(settings): select supported LLM providers · feat: add linux desktop installer build (CI + deb maintainer fix) · feat(automations): show a run's live phase
feat(sidebar): add getting started checklist with settings toggle · feat: show workspace path in Files view · feat(automations): install a catalog entry that ships a script bundle
feat(canvas): present structured error outcomes · feat: add LLM pre-flight validation to prevent saving misconfigured profiles · feat(automations): add a Git Sync page
feat(conversations): add client-side conversation archive · feat(chat): add inline markdown artifact previews · feat: add ready-for-dev issue readiness gate with type-specific criteria
feat: clarify free OpenHands model endpoints
feat: show per-run LLM cost in the Activity Log and exports · feat: reorder Customize navigation · feat(settings): polish Agent Canvas version update UI · feat(home): keep Automations pane visible when empty
feat: set Canvas default model to GLM 5.2 · feat: activity log export · feat: ship the featured automations landing dashboard · feat(skills): filter the skills page with a faceted rail
fix(e2e): disable browser tools during mock LLM builds
fix(backend-registry): pick a healthy local backend as the fallback · fix: update release repository metadata
feat: always show the LLM selector; move agent-profile switching to the tools menu · feat: add persistent agent memory toggle · feat: allow overwriting a secret value from the edit form
fix(release): recover npm publish after tombstoned 1.6.0
revert(agent-profiles): restore default tools in cloud launches
feat: protect Agent Canvas behind SaaS auth
fix(frontend): restore cross-domain PostHog tracking by aligning client/server distinct_id (WIP) · fix(app-server): lower DB pool defaults and make them env-tunable
fix(app-server): preserve conversation created_at across lifecycle webhooks · fix(mcp): preserve SaaS credentials with encrypted storage
feat(app-server): enrich final archive manifests and remove initial snapshots · feat(enterprise): make BYOR key alias pattern configurable
chore(main): release 1.11.0[bot]
feat(enterprise): Agent Profiles on the cloud/SaaS backend (#15044) · feat(backend): Add Budgets dashboard and expand Usage Dashboard · feat: budgets and usage monitoring UI · feat: surface email enabled for smtp/resend
feat: pass repository metadata to observability traces
feat: rename admin dashboard to usage & monitoring · feat: add SMTP email service · feat: track user login timestamps · feat: surface email enabled for smtp/resend
fix(app-server): pass index columns as a list in migration 013 · fix: default ENABLE_ACP on so ACP agent settings show in OH Cloud
feat(enterprise/auth): super-admin management endpoint (grant/revoke/list)
chore(main): release cloud 1.42.0[bot]
feat(enterprise/auth): super-admin management endpoint (grant/revoke/list)
feat: implement semantic file chunking using tree-sitter AST parsing · feat(org): Add organization conversation admin dashboard
feat: move existing users into the default org when first auto-added · feat: route unclaimed automation events to the default org on single-org installs · feat: add hide_personal_workspaces flag for org-only OHE installs
feat: Add full Git history user setting · feat(enterprise): add admin user-provisioning endpoint · feat: Use interrupt endpoint for agent pause UI · feat: add default_sandbox_spec_id to user settings
feat: move existing users into the default org when first auto-added · feat: route unclaimed automation events to the default org on single-org installs · feat: add hide_personal_workspaces flag for org-only OHE installs
Added LLM profiles · Users can select the sandbox grouping strategy · Enabled sub-agent delegation · Minimal generic ACP agent UI
Added support for hooks · Added Catalan language
Added ability to attach or change the Git repository for existing conversations · Added a Task List tab in the right panel that shows the agent's current task list with real-time status updates · Added planning agent
Add MiniMax-M2.5 model support
Added CORS support to Docker sandbox service for remote browser access. Users running OpenHands on remote servers should set WEB_HOST environment variable (e.g., WEB_HOST=192.168.1.100:3000) along with…
Error messages in the chat interface now include a dismiss button and automatically truncate long messages with a 'View More' toggle
Adding a status indicator above the chat until the conversation is ready
202539
Added OAuth 2.0 Device Flow for CLI authentication · Added refresh button to the Changes tab · Added Export Conversation button for conversations in the conversation panel · Added Forgejo integration
New OpenHands 1.0.0 using the new software-agent-sdk with many optimizations across the app. Check out the documentation on how the OpenHands V1 uses the software-agent-sdk.
Cleanup CLI outputs when conversation is first started #11677 · Prevent tool calls with security_risk parameter from crashing conversations #11710
Allow pinning and unpinning of conversation tabs · Add Documentation link to profile menu
Fix broken entrypoint when using UV #11643 · Fix MCP settings persisting even after removal #11602 · Remove preset MCPs from upstream package agent-sdk #11602 · Lazy load conversation when calling /new command #11601
Made terminal tab read-only · Removed Jupyter tab
Allow users to configure settings before conversation has started #11588 · Fix conversation ending abruptly with error when invoking condenser (requires users to re-configure llm settings) #11587
CLI command to launch OpenHands GUI (openhands serve command) #11584
Fixed LLM args to remove deprecation warnings · Fixes from OpenHands SDK dependency for MCP tool validation error
Removed V0 CLI. V0 CLI is officially unsupported
Simplified option to reject agent action in confirmation mode in #11443
CLI binary GLIBC compatibility for older Linux systems · Unexpected/unrecoverable crashes when disabling confirmation mode mid conversation · Existing API key disappearing from CLI settings when updating settings
Multi-platform standalone executable binaries to run CLI without additional setup · Faster startup time running CLI using uv, pip, etc · New and refreshed UI · Simplified JSON MCP configuration + mcp oauth support
Added support for Lemonade Provider
Added configurable timeout settings for SHTTP MCP servers to prevent tool calls from hanging · Added claude-sonnet-4-5 model support
Pinned litellm version to fix dependency issues · Set Claude Sonnet output token limit
New UI for landing and conversation pages · New bring your own key (BYOK) feature in OpenHands Cloud
Added support for AGENTS.md files in microagents · Added new microagent management feature · Added support for routing between different LLMs in an agent trajectory
Added LLM risk analyzer · Added setting for condenser max history size
你是否曾在深夜对着满屏的代码叹息,明明在构建改变世界的产品,却被迫沉迷于永无止境的重复性工作?每一次手动部署、每一次机械的日志检查、每一次枯燥的数据整理,都在悄无声息地吞噬着你宝贵的创造力和时间。这是一个巨大的矛盾:我们拥有前所未有的强大计算能力,却依然被繁琐的流程所束缚。
Added task tracking tool and planning mode for long-horizon tasks · Added a new feedback reason: 'The agent didn't finish the job' · Enabled MCP in OpenHands Cloud
Added support for binding named Docker volumes into the runtime · Allowed partial modification of CLI settings · Added ability to specify the user OpenHands should commit as
Added qwen-3-coder-480b to OpenHands provider · Added gpt-5-2025-08-07 to OpenHands provider · Added support for model claude-opus-4-1-20250805 · Added search API key settings to CLI
Added Moonshot AI Kimi-K2 model to recommended models · Added kimi-k2-0711-preview model to OpenHands provider · Added new options to the CLI confirmation dialogs
Added CLI/VSCode integration · Added OpenHands provider for LLM through OpenHands Cloud · Introduced new Memory UI feature · Added branch name and git provider on conversation cards
Added support for reading microagents from the user directory (/.openhands/microagents/) in addition to existing sources · Added support for .cursorrules. It can be placed in the root directory
Add new feedback options · Add GitLab alternative directory support (openhands-config) for microagents. See updated documentation
Allowed custom model names during CLI model selection
Kubernetes Runtime Support · File and Image Upload Support in conversations · Bitbucket Integration for local usage · Budget Configuration per conversation
想象一下:深夜加班调试代码时,你突然发现团队提交的PR引入了致命漏洞。手动审查数百行变更?Git历史回溯?这场景让开发者脊背发凉。OpenHands正是打破这种困境的钥匙——当传统协作工具在复杂项目中捉襟见肘时,它通过AI驱动自动化将代码审查耗时缩短80%。更震撼的是矛盾点:开发者渴望高效协作,却被碎片化工具链束缚;企业追求交付速度,却因沟通断层付出高昂返工成本。OpenHands用智能工作流缝合这些裂痕,让代码协作像呼吸般自然。
在AI开发领域,开发者常陷于两难困境:要么选择封闭的"黑箱"系统牺牲透明度,要么投入数月搭建基础架构。OpenHands撕开了这个死循环——它像手术刀般精准切中开发者痛点,用开源协作的锋芒刺破效率困局。当某医疗AI团队用传统工具耗时3周仍无法定位模型偏差时,OpenHands的实时协作图谱让20名开发者36小时内完成问题溯源,这个故事正在硅谷技术圈疯传。
在代码的海洋中,开发者们正面临着一场无声的战争。每天有超过62%的工程师在GitHub上遭遇协作困境,34%的项目因沟通断层导致延期。当传统开发工具还在用冰冷的命令行和碎片化的沟通界面割裂团队时,OpenHands像一柄手术刀,精准切开技术协作的毒瘤。这不是又一个平庸的协作平台,而是用AI重构的开发者神经中枢——它能记住每个开发者独特的代码指纹,在提交冲突爆发前预判矛盾点,甚至能自动生成带有情感温度的技术文档…
当你的代码在凌晨三点突然崩溃,当团队协作变成无休止的邮件战争,当重复性工作蚕食着你的创造力——这就是开发者世界的暗面。OpenHands像一柄光剑刺破黑暗,它让AI成为你的第二双手,在自动化测试与智能审查中重构开发秩序。当其他工具还在用代码片段折磨你时,它已悄然将整个开发流程装进了神经网络,让每个commit都闪耀着集体智慧的光辉。这不是工具迭代,而是开发文明的跃迁。
在当今快速发展的科技时代,人工智能的应用已经渗透到我们生活的方方面面。然而,许多开发者在使用AI工具时常常面临着复杂的操作流程和繁琐的设置。这就像在一片茂密的森林中迷失方向,虽然前方有光明的出口,却难以找到通往那里的路径。OpenHands的出现,正是为了打破这一矛盾。它不仅简化了与AI交互的过程,还提供了强大的功能,帮助开发者更高效地实现他们的创意。使用OpenHands,您将能够轻松驾驭AI的力量,释放无限的创造潜能。
在当今这个信息爆炸的时代,如何高效地管理和处理海量数据成为了一个亟待解决的难题。OpenHands应运而生,旨在为开发者提供一个强大的工具,帮助他们在复杂的环境中轻松导航。然而,许多开发者在选择工具时面临着一个矛盾:是选择一个功能强大但复杂的工具,还是一个简单易用但功能有限的工具?OpenHands的出现,正是为了打破这一矛盾,它不仅提供了强大的功能,还保持了用户友好的界面,让开发者能够在高效与易用之间找到完美的平衡。
在这个快速发展的技术时代,开发者们面临着前所未有的挑战。传统的工具往往无法满足日益复杂的需求,导致效率低下和项目延误。而OpenHands的出现,正是为了打破这种困境。它不仅提供了强大的功能,还能有效地简化开发流程,帮助团队更快地实现目标。然而,许多人仍然对它的真正价值感到困惑,甚至怀疑是否值得投入时间去学习和使用。正是在这种矛盾中,OpenHands展现出了它的独特魅力。
在当今快速发展的科技时代,人工智能的应用已经渗透到我们生活的方方面面。然而,许多开发者在构建AI应用时,常常面临着复杂的技术挑战和资源限制。OpenHands的出现,正是为了打破这种困境。它不仅提供了一个强大的平台,帮助开发者更轻松地实现他们的创意,还能有效地提升工作效率。想象一下,一个开发者在面对繁琐的代码时,突然发现了一个能够简化流程的工具,这种矛盾的解决无疑会让他们倍感振奋…
在这个快速变化的科技时代,开发者们面临着越来越多的挑战。传统的工具往往无法满足日益复杂的需求,导致开发效率低下,甚至影响项目的成功。而OpenHands的出现,正是为了解决这一矛盾。它不仅提供了灵活的功能,还能帮助开发者更高效地管理和定制他们的项目。想象一下,如果你能在一个平台上轻松实现个性化的开发体验,提升团队的协作效率,那将是多么令人振奋的事情!
Undated1
在这个AI技术日新月异的时代,开发者们正深陷"创新困境"的泥潭:一边是开源社区海量的AI模型如星辰般璀璨,另一边却是将这些技术落地时面临的"最后一公里"魔咒。OpenHands就像一把瑞士军刀,精准刺破了这个行业痛点——它让原本需要数周才能完成的AI应用集成,缩短到喝杯咖啡的时间。当其他框架还在让开发者与复杂的API文档搏斗时,OpenHands已经搭建好了可插拔的智能组件库,这不仅是效率的革命,更是创造力的解放。