AIGCLISTAIGCLIST
Mastra
AI 工具评分卡

Mastra

一个TypeScript AI代理框架，结合了CLI脚手架、用于快速原型设计的可视化Studio、持久的多步骤工作流以及面向跨职能团队的Agent Builder——全部免费自托管。

免费增值AI 智能体开发mastra.ai
访问
发布于 2026年7月6日

基准评分

Mastra 在 Agent 就绪度与 AI 可见性上的得分 AI 就绪度和 GEO Score 是 VibeLaunch 在提交后生成的平台评估。

由 AIGC List 基准评分提供支持

决策摘要

TypeScript developers and product teams with both engineering and non-engineering collaborators.

Building, prototyping, and deploying AI agents with durable multi-step workflows.

适合

  • TypeScript teams building production AI agents
  • Organizations requiring self-hosted AI infrastructure with data-residency guarantees
  • Cross-functional teams where non-developers need to assemble and publish agents

注意

  • Enterprise pricing requires contacting sales for a custom quote
  • TypeScript-only ecosystem — no Python or multi-language support
  • Requires external model provider API keys; framework does not include model inference

概述

Mastra 是一个强大的 TypeScript AI 框架，旨在简化复杂 AI 智能体和工作流的开发与生产化。它由 Gatsby 背后的团队打造，利用现代 JavaScript 技术栈，赋能开发者轻松构建智能应用。Mastra 旨在简化从原型设计到部署的整个 AI 开发生命周期，为创建能够执行任务、访问数据并保持持久记忆的智能体提供了一套全面的工具。\n\n该框架的脱颖而出之处在于为各种 AI 提供商提供了统一的 API，允许在不同模型之间无缝切换。它强调稳健的工作流编排，具有基于图的状态机、内置追踪和人机回环（human-in-the-loop）能力等特性。对于开发者而言，Mastra 提供了直观的开发者体验 (DX)，并专注于性能、可观测性和可扩展性。\n\n### 核心能力\n- AI 智能体：构建能够执行任务、管理内存和集成工具的智能智能体。\n- 工作流编排：使用具有清晰控制流的持久化、基于图的状态机来设计和管理复杂的 AI 任务序列。\n- RAG 集成：通过同步数据、抓取网页并将其嵌入知识库以进行高效查询，为智能体配备上下文。\n- ���观测性：通过������的��踪、指标和评估工具监控智能体和工作流的性能。\n- 统一 AI 提供商 API：只需更改一行代码即可轻松切换不同的 LLM 提供商。\n\n### 适用人群\nMastra 非常适合希望将先进 AI 能力集成到 JavaScript 和 TypeScript 应用程序中的开发者和团队。无论您是在构建聊天机器人、自动化复杂流程，还是创建数据驱动的 AI 功能，Mastra 都能提供高效开发和可扩展部署所需的工具和结构。它对开发者体验的关注和全面的功能使其成为初创公司和成熟企业的宝贵资产。\n\n### 快速入门\n开始使用 Mastra 非常简单。您可以使用 npm create mastra@latest 来初始化一个新项目。该框架提供了详尽的文档、模板和示例，指导您构建第一个 AI 智能体或工作流。免费书籍《构建 AI 智能体的原则》(Principles of Building AI Agents) 也是学习基础知识的宝贵资源。

评价 (0)

—0 条评分

还没有评价。成为第一个评价的人！

评分构成

编辑评分由哪些维度构成，每项附判断依据。 AI 就绪度和 GEO Score 是 VibeLaunch 在提交后生成的平台评估。

Information quality

Documentation is well-structured for both human and AI consumption, with SKILL.md files, SOURCE_MAP.json, and an llms.txt index. However, all evidence derives from vendor-controlled sources with no independent third-party review in the source packet.

7.2
依赖场景

Packages ship embedded docs in dist/docs including SKILL.md, references, and source maps (claim-08). The project publishes an llms.txt index and per-page markdown export (claim-11).

Ease of use

One-command CLI scaffolding, auto-detected model provider keys, and Studio prototyping create a low-friction onboarding path. The Agent Builder extends accessibility to non-developers, though the TypeScript-only constraint limits audience breadth.

7.8
依赖场景

CLI scaffolds a complete project with agents, workflows, and tools (claim-01, claim-10). Studio provides UI-free prototyping (claim-02). Model keys are auto-detected from environment variables (claim-03).

Feature depth

The core primitives — agents, workflows, tools, memory — cover the essential agent development surface. Evidence for advanced capabilities such as evals, guardrails, multi-agent orchestration, or RAG pipelines is absent from the source packet.

6.8
建议核验

Durable workflows with typed control flow and Observational Memory are claimed (claim-04). Tool definitions require structured createTool() with zod schemas (claim-05). Programmatic agent.generate() supports headless usage (claim-09).

Workflow fit

Durable multi-step workflows with typed control flow address a real need for production agent pipelines. The Agent Builder enables cross-functional collaboration, though the gap between visual assembly and production deployment is not documented in the available evidence.

7.0
依赖场景

Workflows support durable multi-step processes with typed control flow (claim-04). Agent Builder lets non-developers assemble agents visually (claim-07). Programmatic API supports CI/CD and scripted usage (claim-09).

Reliability

The self-hosted deployment model and enterprise VPC guarantees suggest a reliability-conscious architecture, but all claims are vendor-supplied. No independent benchmarks, incident histories, or third-party production deployment reports exist in the source packet.

6.5
建议核验

Self-hosting available for free; enterprise tier guarantees data stays in customer VPC with no traces, prompts, or outputs leaving the environment (claim-06).

Value

A free self-hosted tier is a genuine differentiator. The enterprise flat-fee model avoids per-trace metering surprises, but the custom-quote requirement adds procurement friction and makes total cost opaque without engaging sales.

7.0
依赖场景

Free self-hosted tier confirmed (claim-06). Enterprise pricing is custom flat annual fee with RBAC, SSO, IAM, and VPC isolation (claim-06).

评分反映可查证的产品资料，不代表实际使用效果保证。

Agent 就绪度

评估 Agent 能否通过产品的官方信息理解产品，并重建一条有文档依据的工作流程。

Automated agent-readiness assessment of https://mastra.ai/: 7 of 22 checks verified across 3 fetched pages. No substantial machine interface is documented — agents can understand and cite the product but not operate it. Absent: agent_tooling_artifacts, api_reference, authentication, request_examples, response_examples, error_documentation.

就绪度维度

评估维度得分
文档质量57
执行结果可验证性0
机器接口20
项目定位清晰度100
资源可发现性100
工作流完整度40

对 Agent 有帮助的部分

  • docs: verified during this run
  • llms txt: verified during this run
  • sitemap: verified during this run
  • quickstart: verified during this run
  • sdk: verified during this run
  • webhooks: verified during this run

Agent 受阻的部分

  • No agent instruction files, code-distribution commands, or named slash-command skills found across fetched pages.
  • No api reference signal matched across 3 fetched pages.
  • No authentication signal matched across 3 fetched pages.
  • No request examples signal matched across 3 fetched pages.
  • No response examples signal matched across 3 fetched pages.
  • No error documentation signal matched across 3 fetched pages.

证据核查

关于该工具的公开声明，每条均标注核验状态与引用来源。

guides/getting-started/quickstart4
mastra.ai已验证核验于 2026年7月16日

Mastra provides a CLI command (`npm create mastra@latest`) that scaffolds a project with agents, workflows, and tools, and supports flags such as `--no-example` and `--template`.

Mastra exposes a Studio web interface for building and prototyping agents without requiring developers to construct a custom UI.

Mastra's unified model router supports more than 1,000 models across providers including OpenAI, Anthropic, and Google, and auto-detects provider API keys from environment variables.

The `create mastra` CLI supports a `--default` flag for non-interactive project creation, generating a boilerplate that includes an index entry point, an agent definition, a tool definition, and a workflow definition.

https://mastra.ai/guides/getting-started/quickstart
https://mastra.ai/docs3
mastra.ai已验证核验于 2026年8月30日

A documentation surface is reachable at https://mastra.ai/docs.

A quick-start / agent-skills documentation page is reachable at https://mastra.ai/guides/getting-started/quickstart.

Agent-native positioning with a concrete operational path: "The docs provide a concrete operational path for AI agents with explicit instructions and setup commands.".

https://mastra.ai/docs
ai-agent-framework3
mastra.ai厂商声明核验于 2026年7月16日

Mastra's unified model router supports more than 1,000 models across providers including OpenAI, Anthropic, and Google, and auto-detects provider API keys from environment variables.

Mastra supports durable multi-step agent workflows with typed control flow, and an Observational Memory primitive that persists conversation context and application state across sessions.

Mastra projects can be self-hosted for free, with an optional enterprise tier that adds RBAC, SSO, IAM, network policy integration, and guarantees data stays within the customer's VPC under a flat annual fee.

https://mastra.ai/ai-agent-framework
docs/getting-started/manual-install3
mastra.ai已验证核验于 2026年7月16日

Tools must be defined via the `createTool()` factory from `@mastra/core/tools` with id, description, inputSchema (zod), and execute; plain object tool definitions silently fail to execute.

Agents can be invoked programmatically via `agent.generate()` without launching Studio, suitable for console-based scripts and automated pipelines.

Supported model providers include OpenAI, Anthropic, and Google Gemini, each configurable via environment variables such as `OPENAI_API_KEY`, `ANTHROPIC_API_KEY`, or `GOOGLE_API_KEY`.

https://mastra.ai/docs/getting-started/manual-install
docs/getting-started/build-with-ai2
mastra.ai已验证核验于 2026年7月16日

Mastra packages ship with embedded AI-readable documentation in `dist/docs`, including SKILL.md files, reference documentation, and a SOURCE_MAP.json linking public exports to node_modules locations.

Mastra publishes an `llms.txt` documentation index and provides per-page 'Copy markdown' functionality, making its documentation programmatically accessible to AI coding assistants.

https://mastra.ai/docs/getting-started/build-with-ai
TypeScript AI Framework for Agents and Apps | Mastra1
mastra.ai已验证核验于 2026年8月30日

The entry page was fetched and analyzed for machine-interface signals (title, headings, developer links, keyword probes).

https://mastra.ai/
https://mastra.ai/llms.txt1
mastra.ai已验证核验于 2026年8月30日

llms.txt is published at the site root and readable.

https://mastra.ai/llms.txt
https://mastra.ai/sitemap.xml1
mastra.ai已验证核验于 2026年8月30日

sitemap.xml is reachable and lists site pages.

https://mastra.ai/sitemap.xml
agent-builder1
mastra.ai厂商声明核验于 2026年7月16日

The Agent Builder allows non-developer team members — product managers, operations, and support staff — to assemble tools, models, and workflows into agents they can chat with and publish.

https://mastra.ai/agent-builder

决策核对台

在依赖该产品或访问官网前，最值得先确认的问题。

Mastra is a TypeScript AI agent framework that provides scaffolding, prototyping, and deployment primitives — Agents, Workflows, Tools, and Observational Memory — so teams can build production AI agents without assembling these components from separate libraries.

Run `npm create mastra@latest <project-name>` to scaffold a project with example agents, workflows, and tools. Supply a model provider API key as an environment variable, then launch Studio or call `agent.generate()` to interact with your agent.

Yes. The Agent Builder provides a visual interface where product managers, operations staff, and support personnel can assemble tools, models, and workflows into agents they can test and publish without writing code.

Mastra's unified model router claims support for 1,000+ models. Confirmed providers in the documentation include OpenAI, Anthropic, and Google Gemini, each configured via standard environment variables like `OPENAI_API_KEY` or `ANTHROPIC_API_KEY`.

Mastra projects can be self-hosted for free. An enterprise tier is available with RBAC, SSO, IAM, VPC data isolation, and flat-fee annual pricing — you must contact sales for a custom quote.

Mastra's Observational Memory primitive persists conversation context and application state across sessions. Combined with durable typed workflows, this enables agents that maintain coherence over long-running, multi-step tasks.

请在官网核验

继续探索

相近任务的不同路径

这些工具以带有明确编辑理由的替代关系关联到当前产品。

01Genspark.ai

Genspark.ai

Focuses on search-augmented agent workflows with real-time information retrieval, offering a complementary approach to Mastra's general-purpose agent framework for use cases where live web data is central.

查看档案
02Girikon.AI

Girikon.AI

An enterprise AI orchestration platform with a different deployment and pricing model; worth evaluating for organizations that prefer a managed cloud service over Mastra's self-host-first approach.

查看档案
03Moltbot

Moltbot

An open-source agent runtime that shares Mastra's self-hosted philosophy but targets a different integration surface; useful for teams evaluating open-source alternatives to the free self-hosted tier.

查看档案
查看 Mastra 的全部替代工具