К основному содержанию
плагин

postman — плагин для Claude Code: API из файлов и Context Graph

Плагин Postman для Claude: агент проектирует, мокает, тестирует и документирует API через файлы в репозитории и CLI-команды.

Установить в Claude Code · Через маркетплейс плагинов Все способы и платформы ↓
/plugin install postman@claude-plugins-official
/plugin marketplace add anthropics/claude-plugins-official
Открыть на GitHub

Что делает postman

Плагин даёт агенту доступ к API-разработке через файлы. Коллекции лежат деревом в postman/collections/, где каждый запрос, папка и сохранённый пример — отдельный YAML-файл по схеме v3. Окружения устроены так же, в postman/environments/. Поэтому агент правит один запрос, а не переписывает весь экспорт коллекции, и вы получаете маленький diff в Git.

Схемы описаны для HTTP, GraphQL, gRPC, WebSocket, Socket.IO, MQTT, MCP и LLM-запросов. Из спецификации API агент генерирует запросы и примеры, а линт и тесты можно прогнать локально до того, как что-то уедет в рабочее пространство Postman.

Отдельная часть — Context Graph: приватная карта API-экосистемы организации, которая сводит сигналы из рабочих пространств Postman, GitHub и New Relic. Скилл api-discovery позволяет спросить у графа, что сломается от изменения, ещё до правок кода. Учтите: граф обновляется раз в сутки, поэтому свежие деплои в нём могут отсутствовать — для проверки прямо перед релизом это ограничение. Установщик npx @postman/postman-plugin находит совместимые агенты на машине и настраивает их; для Codex он ещё и одобряет session-start хук, а если API хуков недоступен, установка идёт дальше с пометкой одобрить хук внутри Codex.

Какие задачи решает

  • сгенерировать запросы и примеры прямо из спецификации API
  • замокать эндпоинт до того, как он реализован
  • прогнать линт и тесты коллекции локально перед публикацией
  • получить reviewable diff вместо гигантского экспорта коллекции
  • выяснить, какие сервисы пострадают от изменения эндпоинта

Примеры запросов агенту

  • Собери коллекцию запросов из спецификации openapi.yaml в postman/collections/
  • Подними мок для эндпоинта оплаты, пока бэкенд не готов
  • Спроси у Context Graph, что сломается, если мы поменяем billing API

Как установить postman

Через маркетплейс плагинов

/plugin install postman@claude-plugins-official
/plugin marketplace add anthropics/claude-plugins-official

По шагам, с командами для каждой платформы: Плагины Claude Code: маркетплейсы, установка и как собрать свой

Вопросы о postman

Как установить postman в Claude Code?

Сначала добавьте официальный маркетплейс Anthropic: /plugin marketplace add anthropics/claude-plugins-official. Затем установите плагин: /plugin install postman@claude-plugins-official.

Что делает плагин postman?

Он открывает агенту API-разработку через файлы: спецификации, коллекции, окружения, примеры, моки, документацию и Flows лежат рядом с кодом приложения. Плюс Context Graph подсказывает зависимости и владельцев сервисов.

Что нужно для работы Context Graph?

Он приватный и требует аутентификации. Данные в него стекаются из рабочих пространств Postman, GitHub и New Relic, а обновляется карта раз в сутки.

Исходный текст

SKILL.md · Apache-2.0
<div align="center">

<a href="https://www.postman.com/"><img src="https://assets.getpostman.com/common-share/postman-logo-horizontal-320x132.png" alt="Postman" width="240" /></a>

# Postman Plugin
**Powering API engineering for agents**

The Postman plugin brings filesystem-first API development and organization-wide
API context to coding agents. It enables agents to design, mock, test, document,
monitor, and ship APIs directly from Claude Code, Cursor, and Codex. Every
operation produces inspectable files or CLI commands that fit naturally into
Git and CI, while the Postman Context Graph helps agents understand
dependencies, ownership, runtime behavior, and the likely impact of a change.

[Install](#install) · [Highlights](#highlights)

</div>

## Install

Install Postman in every compatible coding agent detected on your machine:

```bash
npx @postman/postman-plugin
```

One command configures **Antigravity, Claude Code, Codex, GitHub Copilot CLI, Cursor,
Factory Droid, Kimi Code, OpenCode and Pi**. Run it again to update, `status` to see what's installed,
and `remove` to uninstall; `--agent <id>` limits any of them to one agent.

For Codex, installation also approves Postman's session-start hook, which adds
instructions for using Postman's API skills. There is no separate hook prompt
or dump of its instructions. Already trusted hooks are left alone. If Codex's
hook API is unavailable, installation continues with a note to approve the hook
inside Codex. Run the installer and Codex with the same `HOME` and `CODEX_HOME`;
hook approval applies to that profile. If Codex later refreshes the plugin from
`main` and its contents change, run the installer again to approve the update.

You can also use the following commands to install individually:

### Antigravity

```bash
agy plugin install https://github.com/postmanlabs/postman-plugin
```

Run the same command again to update; `agy plugin uninstall postman` removes it.
In the Antigravity app, **Customizations → Plugins → + Install from URL** takes
the same URL.

### Claude Code

[View Postman on Claude Plugins](https://claude.com/plugins/postman)

```bash
claude plugin marketplace add anthropics/claude-plugins-official
claude plugin install postman@claude-plugins-official
```

The first command registers Anthropic's official marketplace, which a fresh
Claude Code doesn't have until an interactive session gets past sign-in. It
does nothing where the marketplace is already registered.

### Codex

[View Postman on ChatGPT Plugins](https://chatgpt.com/plugins/postman?open_in_app)

```bash
codex plugin marketplace add postmanlabs/postman-plugin --ref main
codex plugin marketplace upgrade postman
codex plugin add postman@postman
```

These commands track `main`; review the session-start hook in Codex after a
manual install. The npx installer also approves the hook after each install or
update.

### Cursor

[View Postman on the Cursor Marketplace](https://cursor.com/marketplace/postman)

```text
/add-plugin postman
```

### GitHub Copilot CLI

```bash
copilot plugin marketplace add postmanlabs/postman-plugin
copilot plugin install postman@postman
```

`copilot plugin marketplace update postman`, then `copilot plugin update postman`, updates it. Sign in to Postman's MCP server
with `/mcp` inside a Copilot CLI session.

### Factory Droid

```bash
droid plugin marketplace add https://github.com/postmanlabs/postman-plugin.git
droid plugin install postman@postman-plugin --scope user
```

`droid plugin marketplace update postman-plugin`, then
`droid plugin update postman@postman-plugin --scope user`, updates it. Sign in
to Postman's MCP server with `/mcp` inside a Droid session.

### Kimi Code

Inside a Kimi Code session:

```text
/plugins install https://github.com/postmanlabs/postman-plugin/tree/main
```

Then run `/new` to start a session with the plugin. Run the same command again,
then `/new`, to update; `/plugins remove postman` removes it. Sign in to
Postman's MCP server with `/mcp-config login plugin-postman:postman`.

### OpenCode

```bash
opencode plugin add github:postmanlabs/postman-plugin
```

That is OpenCode 2.0.4 or later. On OpenCode 1.14.33 or later the command is
`opencode plugin --global github:postmanlabs/postman-plugin`.
[opencode/README.md](opencode/README.md#install) has how to update and remove
it, and the older clone-and-loader install.

### Pi

[View Postman in Pi's package gallery](https://pi.dev/packages/@postman/postman-plugin)

```bash
pi install npm:@postman/postman-plugin
```

`pi update npm:@postman/postman-plugin` updates it. Sign in to Postman's MCP
server with `/mcp login postman` inside a Pi session. The shell's `pi mcp login`
doesn't load extensions, so it reports no server named `postman`.

## Highlights

### Filesystem-first API development

All postman resources have a filesystem representation, so your agent can work
with the API ecosystem through the interface it understands best: files. API
specifications, collections, environments, examples, mocks, documentation,
and Flows can live beside the application code.

The git-native [v3 collection schema](skills/collection-schema-v3/) makes this
especially agent-friendly. A collection is a directory tree under
`postman/collections/`, where every request, folder definition, and saved
example is its own YAML file. Environments use the same file-first model under
`postman/environments/`. HTTP, GraphQL, gRPC, WebSocket, Socket.IO, MQTT, MCP,
and LLM requests all have defined schemas the agent can follow.

That means the agent can:

- Read or change one request without rewriting a large collection export.
- Generate requests and examples directly from an API specification.
- Produce small, reviewable Git diffs and resolve changes with normal code
  review workflows.
- Lint and test the files locally before anything is shared with a Postman
  workspace.


### Context Graph: know what breaks before you make a change

A repository can show what an endpoint calls, but rarely who calls it, whether
those consumers are active in production, where they are deployed, or which
team owns them. The Context Graph fills that gap with a private, authenticated,
organization-wide map of your API ecosystem.

It reconciles signals from the systems where API knowledge already lives:

- **Postman workspaces:** specifications, collections, monitors, and mocks
- **GitHub:** repositories, API definitions, and source-level call sites
- **New Relic:** deployments, runtime traffic, latency, errors, and telemetry

The [`api-discovery`](skills/api-discovery/) skill lets the agent start with the
thing you plan to change and ask one natural-language question:

```bash
postman context-graph ask "What could break if we change the billing API?" --wait
```


The graph discovers the surrounding scope—including repositories that are not
checked out locally—before the agent starts editing code. It refreshes nightly
as services, deployments, ownership, and runtime relationships change.

In Postman's controlled benchmark across 468 repositories, starting with this
map used **up to 74% fewer tokens, 52% fewer tool calls, and 72% lower cost**.
Accuracy also improved in 18 of 21 scored prompt-model pairs. Most graph
queries completed in roughly 20–40 seconds. Read the methodology and results in
[Introducing the Context Graph API: One Map of Your API Ecosystem](https://blog.postman.com/introducing-the-context-graph-api-one-map-of-your-api-ecosystem/).

### File-first API mocks

The [`api-mocking`](skills/api-mocking/) skill creates a working mock from an
OpenAPI specification or collection and stores the implementation beside the
API code. The agent can run it locally, add success and failure scenarios, and
test consumers without waiting for the real service to be ready or available.

The mock stays local until you choose to push and deploy it. When teammates or
external systems need access, the same mock can become a durable hosted URL
without rebuilding it in another tool. Pushed to the service's own workspace,
it's the mock other teams' simulations pull first.

### Develop against simulated dependencies

The [`api-simulation`](skills/api-simulation/) skill runs your code (a service
you're changing, a frontend, a test suite or a demo) for real against stand-ins
for the HTTP services it calls, served together on one port from one
`.sim.yaml`. Each stand-in is the owning team's mock when they publish one.
It reuses an existing `.sim.yaml`, proves a change with a request through the
changed path, and answers "what happens if X goes down" by running each failure.
Pushing the simulation is optional and needs your approval. It needs Postman CLI
1.70.0 or later:

```bash
postman simulation run postman/simulations/orders-dev.sim.yaml --port 4900
```

The service under test is never mocked. It runs for real, so a passing run
shows the change works against what the mocks model; it doesn't replace a run
against the real upstreams. The simulation can also run in CI, and any
dependency can take injected latency, error, rate-limit, or chaos conditions to
exercise failure paths.


## How Postman builds context

`postman init` is idempotent and safe to run in any project. Run it before API
engineering work to create the relevant local directories and make Postman
skills available to guide the task, even if the project is already initialized.
See [bootstrap](skills/bootstrap/SKILL.md#start-with-postman-init) for setup
options and workspace binding.

Work done in a Postman folder produces artifacts: collection runs, lint
results, request history, mock deployments. When that folder is connected to a
Postman workspace after `postman init`, those artifacts are published to it as
they are produced. The workspace persists that context for future work and
enables teammates, reviewers, and other agents to share and build on what the
agent did.

What each command contributes to the connected workspace:

| Command | What it publishes |
| --- | --- |
| `postman collection run` | Run results and run history |
| `postman application test` | Test run results |
| `postman spec lint` | Lint results, including violation counts and pass/fail |
| `postman workspace push` | Local workspace contents, persisted as context for future work and sharing with others |
| `postman runner start` | Runner processing and health activity |
| `postman flows run` | Flow run results |
| `postman request` | Agent's API interaction activity |


Publishing happens by default. To change this workflow, you can set 
`--report-events=false` per invocation.

### Anonymous usage counts

The CLI records command usage surface — a minimal,
unauthenticated count with no workspace, user, or payload attached.

### The MCP fallback

The plugin registers Postman's hosted MCP server as a fallback for when the CLI
cannot run. MCP tool calls reach Postman by design and are not governed by the
CLI flags above; skip installing the MCP server if you do not want that path.

## License

Apache-2.0 — see [LICENSE](LICENSE).

© postmanlabs · Apache-2.0 · postmanlabs/postman-plugin

Похожие плагины

Все плагины по теме «Бэкенд и API» →
typescript-lsp официальный Anthropic Подключает языковой сервер для TypeScript и JavaScript к Claude Code. Помогает переходить к определениям, искать ссылки и проверять ошибки прямо при редактировании кода. ★ 44 635звёзд репо ruby-lsp официальный Anthropic Плагин подключает языковой сервер Ruby к Claude Code для анализа кода и навигации по файлам .rb, .rake, .gemspec, .ru и .erb. ★ 44 635звёзд репо php-lsp официальный Anthropic Плагин подключает языковой сервер Intelephense к Claude Code. Он передаёт агенту диагностику и подсказки по коду для файлов с расширением .php. ★ 44 635звёзд репо mcp-server-dev официальный Anthropic Три скилла, которые проведут от вопроса «что я вообще подключаю» до готового MCP-сервера с инструментами, авторизацией и, если нужно, виджетами в чате. ★ 44 635звёзд репо laravel-boost официальный Laravel MCP-сервер с инструментами Laravel: Artisan, Eloquent, роутинг, миграции и генерация кода под фреймворк. ★ 44 635звёзд репо firebase официальный Google Подключает Firebase к Claude через MCP: агент работает с Firestore, аутентификацией, облачными функциями, hosting и storage прямо из проекта. ★ 44 635звёзд репо

Спонсоры проекта

Компании, которые помогают держать аналитику открытой