项目介绍
Agent 开发框架 · 2026-11-09 更新 · 本站尚未实测
Tale 是一个面向团队与 AI 智能体的开源项目工作区(workspace)平台。按照 README 的定位,它把“公司里的问题”转化为任务,交给团队成员与 AI 智能体共同处理:在看板上添加任务、指派人员或智能体、跟踪进展,并查看智能体提交的报告与交付文件。任务说明、讨论、项目知识与最终结果集中保存在同一工作区内。
需要说明的是,Tale 本身是一个平台型产品(平台/工作区),不是单一可下载的智能体或工作流文件。仓库中确实包含智能体运行时支持、工作流自动化编辑器等功能,但使用方式是部署或连接其平台,而非下载某个 Agent 包直接运行。仓库主体语言为 TypeScript,README 显示采用 MIT 许可证,未标明已归档或停止维护。
核心能力
点击能力名称查看输入、输出和原文依据。
根据 README 的“What you can do”列表,平台提供以下功能:
- Chat 对话:在对话中起草、解释与处理信息;可在 Arena 中并排比较两个模型回答,回答引用知识时可查看来源。
- Projects 项目:把相关任务、文件、指令与聊天集中管理,可选择哪些对话与项目共享。
- Project agents 项目智能体:为智能体定义指令、运行时(harness)、模型与工具;指派任务、启动智能体并审阅结果。README 称可配备“经理智能体”,在授予所需工具后把就绪任务委派出去并协调后续工作。
- Knowledge 知识库:为检索准备文档、知识条目与网站内容;上传完成与索引是分开的步骤。
- Automations 自动化:把步骤连成工作流,进行测试,并按手动或配置的触发方式运行。
- Connectors 连接器:用手动管理的凭据连接外部服务。
- Administration 管理:管理成员、角色、模型提供方、策略、用量与审计记录。
README 列出的可用智能体运行时包括 Claude Code、Codex、Cursor、Gemini CLI、Hermes、OpenClaw、OpenCode、Pi 与 Qwen Code,并说明可用性取决于部署方式、凭据与沙箱容量。智能体在持久沙箱工作区中运行,并发数受所配置容量限制。
实际应用场景
适用场景
README 给出的典型用途包括:研究某个问题、审阅文档、准备报告与营销材料,或构建网站、应用与内部工具。示例流程是:创建带问题描述、源文件与完成标准的任务 → 配置项目智能体并选择工具 → 启动任务 → 在看板上跟进、通过任务评论中的 @提及 引导智能体 → 检查其报告与文件 → 对需要重复执行的流程使用带版本管理的自动化。
README 还给出一个“审阅上线简报”的示意任务:比较附件中的简报与会议记录,生成 Markdown 报告,列出冲突日期、缺失负责人与待定决策,并为每项发现标注来源文件与段落,同时保持源文件不变。README 提醒,接受结果前应打开交付报告、对照两个文件核验引用,并确认各类别均已覆盖。
安装与接入
README 提供两条本地路径:
- 本地实例(CLI):README 说明可在 macOS 或 Linux 上使用已发布的 CLI,无需克隆仓库或安装 Bun,并给出如下命令:
curl -fsSL https://raw.githubusercontent.com/tale-project/tale/main/scripts/install-cli.sh | bash
tale init my-project
cd my-project
tale dev
前置条件按 README 为:Docker 与 Compose、为数 GB 的镜像与数据预留空间,以及用于首次回复的模型提供方凭据。Apple Silicon 上,Docker Desktop 自带内置对象存储所需的 amd64 模拟;ARM64 Linux 需在启动前配置模拟。启动后打开打印出的网址,创建首个账户与组织,再在 Settings > AI providers 添加凭据。按 Ctrl-C 停止,在同一目录运行 tale dev 可带着数据恢复。README 称其安装快速入门文档还涵盖 Windows、证书、架构要求与恢复,以及生产部署准备。
- 从源码开发:使用 package.json 中指定的 Bun 版本、兼容的 Node.js 运行时,以及为后端服务准备的 Docker;完整仓库检查还需 Python 与 uv。
bun install --frozen-lockfile
bun run setup:check
bun run dev
README 提示 setup check 仅覆盖部分环境,通过检查不等于数据库、存储或模型提供方已配置完成。
获取方式上,可选择自托管(自有基础设施)或使用其托管 Cloud 服务;README 也提到可通过“请求演示”获取托管实例,并提供 API 入门文档用于连接其他应用。
使用条件与限制
- 是否需要部署
- 需自行准备运行环境,具体要求见正文
- 是否需要注册
- 账号和认证要求见正文;未明确说明的项目仍待核验
- 是否需要 API Key
- 待核验
- 费用与额度
- 费用边界见正文,项目许可不代表模型或托管服务免费
- 所需权限
- 待核验
- 许可证与服务条款
- 待核验
权限费用与限制
- 凭据与计费:可使用自有提供方 API Key 或受支持的厂商订阅搭配兼容的智能体运行时。README 说明,存储的 API Key 经由 Tale 的模型网关(model gateway);受支持的厂商订阅仅在兼容运行时下可用,不能驱动普通 Chat,且会绕过 Tale 网关的计量与支出上限,因此建议先阅读其凭据与成本说明再选择连接方式。
- 版本与费用:README 称 Community 与 Enterprise 包含相同的产品功能,代码以 MIT 许可证发布;Enterprise 增加专业运维与支持。当前服务条款与价格需查阅其官方的 plans and pricing 页面,具体金额本资料未给出,待核验。
- 数据流向:README 明确应用记录、可检索知识与原始文件有各自的存储设置;模型提供方、连接器与外部工具可能在上述存储之外处理数据,仅靠自托管并不能让所有请求都留在本地。选择部署边界前应查阅数据驻留(data residency)说明。本地模型可使推理留在自有基础设施上,但连接的服务与外部工具仍各有数据流向。
- 功能范围:功能取决于角色与部署配置;沙箱并发受配置容量限制。
国内使用情况
网友核验 用户经验分享
与站长实测独立,不代表官方认证或本站已实测。
最近核验:暂无有效核验。参与人数按当前项目有效身份去重;不同核验项分别统计,不合并成用户总结果。仅统计审核通过且未过期的记录。
分项统计、有效期与样本量
| 核验项目 | 有效人数 | 成功 / 部分 / 失败 / 无法判断 | 成功率 | 最近核验 / 有效期 |
|---|---|---|---|---|
| 官方网站访问 | 0 | 0 / 0 / 0 / 0 | 样本不足 | 暂无 / 7 天 |
| GitHub 仓库访问 | 0 | 0 / 0 / 0 / 0 | 样本不足 | 暂无 / 7 天 |
| 安装与运行 | 0 | 0 / 0 / 0 / 0 | 样本不足 | 暂无 / 30 天 |
| 官方服务国内连接 / MCP 连接 | 0 | 0 / 0 / 0 / 0 | 样本不足 | 暂无 / 30 天 |
| 国内网络情况 | 0 | 0 / 0 / 0 / 0 | 样本不足 | 暂无 / 7 天 |
| 国内模型兼容性 | 0 | 0 / 0 / 0 / 0 | 样本不足 | 暂无 / 30 天 |
| 自定义 API | 0 | 0 / 0 / 0 / 0 | 样本不足 | 暂无 / 30 天 |
| 海外服务依赖 | 0 | 0 / 0 / 0 / 0 | 样本不足 | 暂无 / 30 天 |
| 海外账号要求 | 0 | 0 / 0 / 0 / 0 | 样本不足 | 暂无 / 30 天 |
| 中文界面 | 0 | 0 / 0 / 0 / 0 | 样本不足 | 暂无 / 30 天 |
| 客户端兼容性 | 0 | 0 / 0 / 0 / 0 | 样本不足 | 暂无 / 30 天 |
| 账号及费用信息 | 0 | 0 / 0 / 0 / 0 | 样本不足 | 暂无 / 30 天 |
成功率 = 成功 ÷(成功 + 部分成功 + 失败);无法判断不计入分母。至少 5 份可判断结果才显示成功率。账号、费用、海外依赖与语言等事实项,成功表示核查结论与资料吻合,具体观察另行统计;不表示免费或没有海外依赖。不同网络、客户端和版本的结果不能直接等同。
最近有效核验
暂无审核通过且未过期的网友核验。欢迎提交真实使用记录。
查看全部有效核验 →国内实测状态:国内可用性待核验。协议兼容和官方资料说明不等于国内网络可用。
- 官方网站访问
- 尚未实测 待核验
- GitHub 仓库访问
- 尚未实测 待核验
- 官方服务国内连接
- 尚未实测 待核验
- 国内网络情况
- 尚未实测 待核验
- 国内模型兼容性
- 尚未实测 待核验
- 自定义 API
- 尚未实测 待核验
- 海外服务依赖
- 尚未实测 待核验
- 海外账号要求
- 尚未实测 待核验
- 中文界面
- 尚未实测 待核验
- 国内最后核验时间
- 尚未实测
许可与国内使用
仓库的 LICENSE 文件为 MIT 许可证,README 亦声明代码以 MIT 许可发布。MIT 许可适用于该仓库代码;商业模块、托管 Cloud 服务、第三方模型与第三方连接器的条款需另行核验,公开仓库不等于全部服务免费。
README 明确支持的智能体运行时中包括 Qwen Code,但资料未说明在中国大陆网络环境下的可用性、部署地域或针对国内模型的专门适配结论,上述内容均待核验。README 文档提供英语、德语与法语版本;界面截图显示为英文界面,中文界面支持情况本资料未说明,待核验。
项目资料与来源
字段依据与待核验事项
access:来源确认 · Use the published CLI on macOS or Linux; no repository clone or Bun installation is needed:
project_type:来源确认 · Tale gives teammates and AI agents a shared project workspace.
原始 README(折叠技术参考)
<div align="center">
<picture>
<source media="(prefers-color-scheme: dark)" srcset=".github/assets/logo-dark.svg">
<img alt="Tale" src=".github/assets/logo-light.svg" width="150">
</picture>
[](https://github.com/tale-project/tale/actions/workflows/build.yml)
[](https://github.com/tale-project/tale/actions/workflows/checks.yml)
[](LICENSE)
[English](README.md) · [Deutsch](README.de.md) · [Français](README.fr.md)
</div>
# Tale — The open-source workspace for teams and AI agents
**Turn company problems into tasks your team and AI agents can solve together.**
Tale gives teammates and AI agents a shared project workspace. Add tasks to the board, assign people or agents, follow the work, and review reports and delivered files. Keep the brief, discussions, project knowledge, and results together. Use agents to research a question, review documents, prepare reports and marketing materials, or build a website, app, or internal tool.
Choose each agent’s runtime, model, skills, and tools. Equip a manager agent to delegate ready tasks and coordinate follow-up work. Agents run in persistent sandbox workspaces, with concurrency limited by your configured capacity.
Use your own provider API keys or supported subscriptions with compatible agent runtimes. See [runtime and credential support](https://docs.tale.dev/platform/agents/harnesses) for the available combinations.
Self-host Tale on your own infrastructure or use the managed Cloud service. The code is MIT-licensed. Community and Enterprise include the same product features; Enterprise adds professional operation and support. See [plans and pricing](https://tale.dev/pricing) for the current offer.
<table>
<tr>
<td width="50%" valign="top">
<a href="services/docs/public/images/platform/projects-task-board.webp"><img src=".github/assets/readme-gallery-tasks.webp" alt="The Website relaunch task board groups cards by status." width="100%"></a>
<br><a href="https://docs.tale.dev/platform/projects/tasks"><b>Project tasks</b></a><br><sub>Organize work and review its progress.</sub>
</td>
<td width="50%" valign="top">
<a href="services/docs/public/images/platform/project-agents-models.webp"><img src=".github/assets/readme-gallery-project-agents.webp" alt="The project Agents tab lists named agents with their runtime and model." width="100%"></a>
<br><a href="https://docs.tale.dev/platform/projects/project-agents"><b>Project agents</b></a><br><sub>Choose instructions, runtime, model and tools.</sub>
</td>
</tr>
<tr>
<td width="50%" valign="top">
<a href="services/docs/public/images/platform/automation-editor-canvas.webp"><img src=".github/assets/readme-gallery-workflow-editor.webp" alt="The automation editor shows connected steps and the selected node’s settings." width="100%"></a>
<br><a href="https://docs.tale.dev/platform/automations/editor"><b>Workflow editor</b></a><br><sub>Inspect steps, test inputs and review runs.</sub>
</td>
<td width="50%" valign="top">
<a href="services/docs/public/images/platform/chat-arena-split.webp"><img src=".github/assets/readme-gallery-chat-arena.webp" alt="Arena displays two responses to the same prompt and the voting controls." width="100%"></a>
<br><a href="https://docs.tale.dev/platform/chat/arena-mode"><b>Chat and Arena</b></a><br><sub>Compare two model responses side by side.</sub>
</td>
</tr>
<tr>
<td width="50%" valign="top">
<a href="services/docs/public/images/platform/connectors-add-credential.webp"><img src=".github/assets/readme-gallery-connectors.webp" alt="The Add credential dialog lists available connector integrations." width="100%"></a>
<br><a href="https://docs.tale.dev/platform/connectors/overview"><b>Connectors</b></a><br><sub>Choose the services your workspace uses.</sub>
</td>
<td width="50%" valign="top">
<a href="services/docs/public/images/platform/governance-guardrails.webp"><img src=".github/assets/readme-gallery-guardrails.webp" alt="The Guardrails settings page shows policy status and configuration controls." width="100%"></a>
<br><a href="https://docs.tale.dev/platform/admin/governance/guardrails"><b>Governance</b></a><br><sub>Review content safety and data policies.</sub>
</td>
</tr>
</table>
Select a screenshot to view it at full size. The captures show the English interface.
## From tasks to reviewed results
1. **Describe the work.** Create a [project task](https://docs.tale.dev/platform/projects/tasks) with the problem, source files, and completion criteria.
2. **Assign people and agents.** Configure [project agents](https://docs.tale.dev/platform/projects/project-agents) for the work, choose their tools, and start the tasks you want them to handle.
3. **Coordinate and review.** Follow progress on the board, steer agents with @mentions in task comments, and inspect their reports and files. A [manager agent](https://docs.tale.dev/platform/projects/task-automation#let-a-manager-agent-keep-the-queue-moving) can delegate ready work when granted the required tools.
4. **Repeat a defined process.** Use a versioned [automation](https://docs.tale.dev/platform/automations/concepts) when the work needs scheduled starts, workflow steps, or connector approvals.
### Example: review a launch brief
With a project agent and compatible model credentials configured, adapt this illustrative task brief to your own source files:
> Compare the attached launch brief and meeting notes. Produce a Markdown report listing conflicting dates, missing owners, and open decisions. Cite the source file and passage for each finding. Separate confirmed facts from questions, and leave the source files unchanged.
Before accepting the result, open the delivered report, verify its citations against both files, and check that each requested category is covered. Ask for corrections in the task when evidence is missing. The [task review guide](https://docs.tale.dev/platform/projects/task-automation) explains how to request changes or accept completed work.
## Evaluate Tale for your team
**Which agent runtimes can I use?** Tale includes Claude Code, Codex, Cursor, Gemini CLI, Hermes, OpenClaw, OpenCode, Pi, and Qwen Code. Availability depends on your deployment, credentials, and sandbox capacity. Check the [runtime compatibility matrix](https://docs.tale.dev/platform/agents/harnesses) for credential paths, tools, and conversation limits.
**Can I use an API key or an existing subscription?** Stored provider API keys use Tale's model gateway. Supported vendor subscriptions work only with compatible runtimes, cannot power ordinary Chat, and bypass Tale's gateway metering and spending caps. Read the [credential and cost details](https://docs.tale.dev/platform/agents/harnesses#understand-credential-exposure-and-cost) before choosing a connection.
**What does self-hosting require?** Start with Docker and Compose, storage for images and persistent data, and credentials for a supported model provider. Production also needs DNS, TLS, backups, and access controls. The [self-hosted quickstart](https://docs.tale.dev/self-hosted/install/quickstart) covers the local setup and links to production preparation.
**Where does my data go?** Application records, searchable knowledge, and original files have separate storage settings. Model providers, connectors, and external tools can process data outside those stores; self-hosting alone does not keep every request local. Review [data residency](https://docs.tale.dev/self-hosted/configuration/data-residency) and the [runtime's credential and network behavior](https://docs.tale.dev/platform/agents/harnesses).
**What differs between Community and Enterprise?** Both include the same product features under the MIT license. Enterprise adds professional operation and support. See [plans and pricing](https://tale.dev/pricing) for the current service terms.
## Start here
| Your goal | Follow this guide |
| --- | --- |
| Use a workspace your team already has | [Send your first message](https://docs.tale.dev/get-started/quickstart) |
| Install Tale | [Self-hosted quickstart](https://docs.tale.dev/self-hosted/install/quickstart) |
| Get a managed instance | [Request a demo](https://tale.dev/request-demo) |
| Build an agent for a project | [Create and test a project agent](https://docs.tale.dev/get-started/editors) |
| Connect another application | [API getting started](https://docs.tale.dev/get-started/developers) |
| Change the source code | [Contributor setup](docs/en/develop/contributor-setup.md) |
| Build with Tale’s UI components | [Component guides and live examples](https://ui.tale.dev/docs/getting-started/introduction) |
### Run a local instance
Use the published CLI on macOS or Linux; no repository clone or Bun installation is needed:
```bash
curl -fsSL https://raw.githubusercontent.com/tale-project/tale/main/scripts/install-cli.sh | bash
tale init my-project
cd my-project
tale dev
```
You need Docker with Compose, space for several GB of images and your data, and a model provider credential for your first reply. Docker Desktop includes the amd64 emulation needed by the bundled object store on Apple Silicon. On ARM64 Linux, configure emulation before starting.
The CLI can help install or start Docker. On startup, open the printed URL, create the first account and organization, then add a credential under **Settings > AI providers** and [send your first message](https://docs.tale.dev/get-started/quickstart).
Press `Ctrl-C` to stop; run `tale dev` in the same directory to resume with your data.
The [installation quickstart](https://docs.tale.dev/self-hosted/install/quickstart) covers Windows, certificates, architecture requirements and recovery. Use the [CLI guide](tools/cli/README.md) for commands. Before serving a team, follow [production preparation](https://docs.tale.dev/self-hosted/install/quickstart#prepare-a-production-deployment); `tale deploy` uses separate data volumes from the local development instance.
### Develop from source
Use the Bun version in [package.json](package.json), a compatible Node.js runtime, and Docker for the backing services. Python and uv are also needed for the full repository checks. The [contributor setup guide](docs/en/develop/contributor-setup.md) covers versions, environment variables, and ports.
```bash
bun install --frozen-lockfile
bun run setup:check
bun run dev
```
Wait for the platform’s readiness message, then open the address it prints. The setup check covers only part of the environment; a passing check does not establish that databases, storage, or a model provider are configured.
To work on documentation alone, no platform database or provider is needed. The first command previews the product docs, the second the design-system guide:
```bash
bun run --filter @tale/docs dev
bun run --filter @tale/ui-docs dev
```
## What you can do
- **[Chat](https://docs.tale.dev/platform/chat/basics):** draft, explain, and work with information in a conversation. Compare models in Arena and inspect sources when an answer uses knowledge.
- **[Projects](https://docs.tale.dev/platform/projects/overview):** keep related tasks, files, instructions, and chats together. Choose which chats to share with the project.
- **[Project agents](https://docs.tale.dev/platform/projects/project-agents):** define an agent’s instructions, harness, model, and tools. Assign a task, start the agent, and review the result.
- **[Knowledge](https://docs.tale.dev/platform/knowledge/overview):** prepare documents, knowledge entries, and website content for retrieval. Upload completion and indexing are separate steps.
- **[Automations](https://docs.tale.dev/platform/automations/concepts):** connect steps into a workflow, test it, and run it manually or through configured triggers.
- **[Connectors](https://docs.tale.dev/platform/connectors/overview):** connect external services using credentials your workspace controls.
- **[Administration](https://docs.tale.dev/platform/admin/overview):** manage members, roles, providers, policies, usage, and audit records.
Features depend on your role and deployment configuration. A local model keeps inference on your infrastructure; connected services and external tools still have their own data flows. See [data residency](https://docs.tale.dev/self-hosted/configuration/data-residency) before choosing a deployment boundary.
## Documentation
The documentation is available in [English](https://docs.tale.dev), [German](https://docs.tale.dev/de), and [French](https://docs.tale.dev/fr).
Start with a guided task, use the Platform pages for everyday work, and consult the operator or API references when you need exact configuration details. The [screenshot gallery](SCREENSHOTS.md) gives a visual overview. To improve the docs, read [the docs workspace README](services/docs/README.md).
Building an interface with Tale’s components? The [design-system guide](https://ui.tale.dev) documents `@tale/ui` and `@tale/marketing-ui` with live examples, in English. Its pages live in [services/ui-docs](services/ui-docs/README.md).
## Contribute or get help
Read [CONTRIBUTING.md](.github/CONTRIBUTING.md) and the [repository contract](AGENTS.md) before changing code. Run `bun run check` before submitting a pull request; follow the guide for additional checks appropriate to your change.
- Ask questions in [GitHub Discussions](https://github.com/tale-project/tale/discussions).
- Report reproducible bugs in [GitHub Issues](https://github.com/tale-project/tale/issues), including the version, steps, expected result, and actual result.
- Report vulnerabilities through [private security reporting](https://github.com/tale-project/tale/security).
## License
Tale is available under the [MIT license](LICENSE).
历史前置条件说明(补充参考)
需 Docker 与 Compose、数 GB 镜像与数据空间;首次使用需受支持模型提供方的 API Key 或兼容厂商订阅;从源码开发需 package.json 指定版本的 Bun、兼容 Node.js,完整检查另需 Python 与 uv;生产部署另需 DNS、TLS、备份与访问控制(依 README);托管 Cloud 服务的费用与条款待向官方核验;中国大陆网络与部署地域适配情况待核验。
项目来源与清单版权许可
https://github.com/tale-project/tale
整理来源:https://github.com/slavakurilyak/awesome-ai-agents
清单版权与许可:
MIT License
Copyright (c) 2024 Slava Kurilyak
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.