Postman Flows

Postman Flows is a visual, low-code tool that connects API requests, logic, and AI agents into flows that developers and coding agents can test, run locally, and deploy to Postman Cloud.

始める

エージェント向けにインストール

npm install -g postman-cli
Get started with
Example run with illustrative output: Claude Code in a terminal. Asked to add evals for new AI features, run them every day, and send the results to the team Slack channel, it finds the Support assistant flow, adds correctness and relevance evals to its Summarize ticket and Suggest reply steps, a daily 9:00 UTC schedule, and a Slack step, and deploys the flow to Postman Cloud. The next morning, asked how the evals did, it lists the flow's runs, opens the scheduled 9:00 run, and finds Suggest reply scoring 72, under the 80 threshold, because its prompt hard-codes an old 14-day refund policy. It changes the code to read the policy from the API, so the next scheduled run re-scores it and posts the result to #team-ai.

世界中の導入企業

PayPal社のロゴ
Moneris社のロゴ
Adyen社のロゴ
Notion社のロゴ
Built社のロゴ
AWS社のロゴ

Install the Postman plugin in your coding agent

The Postman plugin gives your agent the flows skill. It lists your flows, runs a flow file against a local environment, deploys it to Postman Cloud, and traces a failed run to the block that broke. It asks before it deploys a flow or changes a live trigger.

get the Postman plugin

ステップ 1
claude plugin marketplace add anthropics/claude-plugins-official
ステップ 2
claude plugin install postman@claude-plugins-official
See all Postman skills

Try these prompts

WHY POSTMAN FLOWS

Agentic where you want it,
deterministic everywhere else

Postman Flows puts API requests, branching logic, AI agents, and evals in one workflow. A model makes the judgment calls you hand it, and fixed blocks handle the rest, so the same flow behaves the same way on your machine, in CI, and as a deployed API.
  • AI agents and fixed logic in one flow

    Put an AI Agent block next to Condition, If, and For blocks. The agent can call MCP server tools or your other flows as tools, while branches, loops, and validation run the same way every time.

  • Evals on every AI step

    Score AI Agent output for correctness, relevance, safety, friendliness, and non-toxicity, or add up to 50 custom evals. Write your own checks in TypeScript in an Evaluate block. Evals run each time the flow runs.

  • Test what happens between endpoints

    Real API journeys branch. Chain requests, map each response into the next step, and use Condition and Validate blocks to test every path, like the out-of-stock order or the expired token.

  • Local runs and cloud runs

    Run a flow on your machine with a local environment file, or in Postman Cloud. Local runs can reach private APIs. Cloud runs use only the shared values of your environment variables, so local values stay on your machine.

  • Deploy as an API, a schedule, or an MCP server

    Deploy any flow to Postman Cloud with a Request, Schedule, or MCP trigger. Each deployment gets a URL, optional authentication, and a snapshot you can redeploy later.

  • A failing flow fails the build

    When a flow fails, the Postman CLI exits with a nonzero code and can write an HTML test report, so CI stops the merge. Running flows from the CLI is available on Enterprise plans.

A Postman Flows eval flow after a run: a Request block feeds an AI Agent block on OpenAI GPT-5 with the Friendliness and Safety evals on. An Evaluate block runs a custom pm.eval check on its result, and a Display block shows the eval scores, passed 2 of 2 at 98%. The Run logs panel below summarises the run: 15.7 seconds, 2 of 2 evals passed, Friendliness 95% and Safety 100%.

USE CASES

What teams build with Postman Flows

Postman Flows takes the multi-step API work that lives in scripts, runbooks, and someone's memory, and makes it something anyone on the team can run.

Grade every agent's answer

When an alert fires, an AI Agent block checks New Relic, triages it as p0, p1, or p2, and pages the on-call engineer through PagerDuty. A second agent suggests a fix and opens a draft PR. Evals grade each agent's answer on every run, so you can see how well it did.

  • AI Agent blocks reach New Relic, PagerDuty, and GitHub through MCP servers
  • Quality presets turn on evals like correctness and relevance
  • Eval Scores show a pass or fail and a score for each eval
  • Each agent can run on a different model
Postman Flows canvas for a Triage with Evals flow: a Start block with service, repo, and team inputs feeds an AI Agent block on Claude Sonnet 4.5 with New Relic and PagerDuty MCP servers and two evals on. Its Eval Scores display shows Correctness and Relevance passing, and a second AI Agent block on GPT-5 with Cursor and GitHub MCP servers takes its result to fix the incident.
Flowsは、開発の進め方を大きく変えてくれました。カスタムスクリプトもつなぎ込みのコードも不要です。複雑なAPIシーケンスを視覚的に設計し、実行状況をリアルタイムで確認できるほか、AIリクエストブロックも直接組み込めます。統合作業にかかる時間を大幅に短縮し、自信を持って開発を進められるようになりました。”
PayPal社のロゴ

PayPal Developer Experienceチーム

PayPal

Read the PayPal customer story

Frequently asked questions

Trusted by over 50 million developers and 500,000 companies.

Install for your agents

npm install -g postman-cli
Get started with

Install for desktop

Sign up for free