开发者工具链概览
2026/8/8 00:27:22 · 来源
AI 翻译于 2026/8/8 00:50:54 · 使用 Qwen3.6 35B (fast, default)
开发者工具链
Palantir 平台附带一套开发者工具,使您能够基于本体(Ontology)进行构建,并通过新的工作流、应用程序和功能释放价值。借助 Palantir 的开发者工具链,您可以直接从开发环境中访问本体的全部功能。
除文中其他地方所述的数据连接和应用程序构建工具外,Palantir 开发者工具链还提供:
- 核心 API 和 SDK,将本体开放给您的应用程序。
- 开发环境和面向开发者的功能,以加速您的工作流。
- Palantir MCP,使 AI IDE 和 AI 代理能够使用 Palantir 上下文并在 Foundry 中执行操作。
- Ontology MCP,将本体资源作为 MCP 工具暴露给外部 AI 代理。
- 计算模块,用于将代码容器化并扩展您的部署规模。
Palantir 开发者服务的核心是开放平台并允许您构建与 Palantir 平台及本体数据直接集成的应用程序的 API 和 SDK。
本体 SDK (OSDK) 使开发人员能够从其本体生成 Python、Java 和 TypeScript 语言的 SDK。OSDK 允许您访问对象类型、应用操作以更新本体中的数据、调用函数,以及为 支持 AIP 的注册运行 AIP Logic 函数。平台内的开发者控制面板包含为您应用程序所选实体的本体特定文档。应用程序通过 OAuth 流程作为公开或机密客户端来访问数据。
Palantir 的 API 通过支持对平台访问和本体底层数据进行编程化管��,助力在 Palantir 平台上构建应用的开发��。API 参考文档也可通过平台内的开发者控制面板获取。
可用的 Palantir API 包括:
有关端点的完整列表(包括模型、审计日志、数据健康监测、流、检查点和 Notepad),请参阅完整的 API 文档。
Palantir 的平台 SDK 公开提供 Python ↗ 和 TypeScript ↗,并随新 API 的推出不断扩充。这些 SDK 可与 Palantir 平台客户端配合,与 Ontology SDK 共同使用。
除了专注于数据连接的开发工具如 Pipeline Builder ↗ 和 Code Repositories ↗ 之外,Code Workspaces ↗ 允许您在基于 Palantir 平台构建应用时使用熟悉的 IDE(如 VS Code)。VS Code 工作区与 Developer Console 集成,使您能够快速 构建 React 应用程序。您可以从平台内的 Developer Console 创建 VS Code 工作区。 Palantir 面向开发者的功能旨在加速平台上的构建流程。Global Branching ↗ 等功能有助于为端到端工作流的开发营造一个安全、高效的环境。
Palantir MCP 是 Model Context Protocol ↗ 的一种实现,它允许 AI IDE 和 AI 代理在 Palantir 平台上自主构建端到端应用程序——涵盖从数据集成、本体(Ontology)配置到应用开发的全过程。此外,您还可以使用 Palantir MCP 使外部 AI 系统能够查询文档、元数据和数据,并在平台上执行高级任务。开发人员可以利用 Palantir MCP 自动化辅助工作,从而专注于核心系统的构建。
Ontology MCP (OMCP) 是 Developer Console 的一项功能,它将应用程序的本体资源暴露为 Model Context Protocol (MCP) 工具,使外部 AI 代理能够与您的本体进行交互。Palantir MCP 主要面向本体构建者和开发工作流,而 Ontology MCP 则主要面向本体消费者:即需要安全读取和写入生产环境本体数据的外部 AI 代理。 Ontology MCP 将对象类型、动作类型和查询功能暴露为 MCP 工具,并可与桌面代理及无头(headless)代理框架集成。
Compute Modules 功能使您能够在 Palantir 平台上部署交互式容器,从而将您的现有代码库(不论使用何种语言)引入平台并在其中运行。借助 Compute Modules,例如您可以将第三方机器学习模型带入平台,并将其集成到您的工作流中。 如需了解更多关于 Compute Modules 的信息,您可以观看官方 Build with AIP: Compute Modules ↗ 视频并参考 文档。
自定义端点
自定义端点 使开发者能够配置并部署用户自定义的 API 端点,支持独立的 URL 模式、请求与响应结构及规范,同时依托 Foundry 的后端能力。这些端点由本体的操作和函数提供底层支撑。
🎙 朗读音色: Dylan (北京话)
学习地图
Learning Map: Palantir Developer Toolchain
Stage 1 — Foundations
- Understand Palantir Foundry's architecture: data, Ontology, and applications
- Datasets, Code Repositories, Pipeline Builder basics
- Onboard to the Developer Console environment
Stage 2 — Core APIs & SDKs
- Use OSDK to generate typed clients (Python/Java/TypeScript) from your Ontology
- Explore platform APIs: Datasets, Filesystem, Administration, Orchestration, Connectivity, SQL Queries
- Install and configure the Python and TypeScript Platform SDKs
Stage 3 — Developer Environments
- Set up Code Workspaces with VS Code integration
- Use Global Branching for safe end-to-end workflow development
- Build React applications through the Developer Console
Stage 4 — AI & MCP Integration
- Deploy Palantir MCP to enable AI IDEs/agents to build and query the platform autonomously
- Expose Ontology resources as tools via Ontology MCP (OMCP) for external agent consumption
- Chain AI actions with ontology reads/writes
Stage 5 — Compute & Customization
- Deploy containerized workloads with Compute Modules for ML models and other code
- Configure custom endpoints on top of the Ontology for user-defined API patterns
Stage 6 — Production Readiness
- Combine all layers: data pipelines → Ontology actions → AI agents → compute → custom APIs
- Design scalable, end-to-end workflows across the full toolchain
动手实践——分步指南
- Create a Palantir Foundry developer account and log into the in-platform Developer Console.
- Navigate to Core APIs & SDKs and generate an OSDK client in your preferred language (Python/Java/TypeScript) from your Ontology data model.
- Install the corresponding Platform SDK:
pip install palantir-foundryfor Python ornpm install @palantir/foundry-platform-typescriptfor TypeScript. - Set up OAuth credentials (public or confidential client) to authenticate API calls against your Palantir organization.
- Open Code Workspaces and connect it to VS Code, following the workspace setup guide in the Developer Console.
- Experiment with a core API endpoint — for example, use the Datasets API to list datasets, read a dataset, or create a new branch.
- Configure Palantir MCP: enable the MCP server in your Foundry environment and connect it to an AI IDE (e.g., Claude Desktop, VS Code Copilot) so your agent can read Ontology metadata and build applications autonomously.
- Optionally expose specific Ontology resources as tools via Ontology MCP (OMCP) so external AI agents can query and write production data safely.
- Package a self-contained piece of code or ML model into a Compute Module to run it inside the platform as an interactive container.
- Define custom endpoints with your own URL patterns and wire them to ontology actions and functions, then test via curl or Postman.
三大推荐资源
- 1Palantir Foundry Developer Toolchain Official Docs
The canonical overview of Palantir's full developer toolchain, covering APIs, SDKs, MCP integration, compute modules, and custom endpoints.
https://www.palantir.com/docs/foundry/dev-toolchain/overview/
- 2Palantir Ontology SDK (OSDK) Documentation
Step-by-step reference for generating typed SDKs from your Ontology in Python, Java, and TypeScript, including OAuth flow and Developer Console docs.
https://www.palantir.com/docs/foundry/ontology-sdk/overview/
- 3Palantir Foundry Python Platform SDK (GitHub)
The official open-source Python SDK for interacting with Palantir's platform APIs, with installation instructions and code examples.
https://github.com/palantir/foundry-platform-python
链接由 AI 推荐——使用前建议快速核实。