How to use Claude to turn web apps into MCP tools
A captured HAR file became OpenAPI documentation, followed by a local Python MCP server that used Playwright to obtain credentials before calling internal APIs.
news.ycombinator.com
Prompt to try
Editorial adaptation, not the source author’s exact prompt. Replace bracketed inputs and review tool permissions first.
Analyze [HAR capture] from [authorized web application] and document the observed endpoints as OpenAPI JSON. Help build a Python stdio MCP server for [limited workflow], with Playwright login detection and local credential caching.
Reach people exploring AI agents
See placements →What happened
The author reported fairly reliable local MCP servers using individual user credentials, including in locked-down corporate environments.
Reported by ksri on Hacker NewsHow the agent did it
- Navigate the web application in Chrome and export network traffic as HAR.
- Ask Claude to analyze the HAR and document APIs as OpenAPI JSON.
- Write a Python script bridging stdio MCP and the internal APIs.
Where a person stepped in
The author directed the limited use case and reported reliability; the post does not describe formal code review.
A person approved each step before the agent continued.
The author wanted Claude Code to connect to third-party web applications that lacked public APIs. These were single-page applications, and the intended integration served a limited use case. They opened Chrome, navigated the entire website, and downloaded the browser network activity as a HAR file. They then asked Claude to analyze that capture and document the APIs as OpenAPI JSON.
Next, they wrote a small Python script with two sides: a standard-input/output MCP interface and calls to the third-party application's internal APIs. Authentication was the remaining piece. When Claude connects, the MCP server launches a Playwright-controlled browser and opens the target application. It detects whether the user is logged in, extracts authentication credentials, saves them in a local cache file, and closes the browser. Subsequent API access runs directly without keeping a browser open.
The author reported producing local MCP servers using each user's credentials in about an hour of work with Claude. They described the result as fairly reliable and said it worked in otherwise locked-down corporate environments.
Who it's for: Developers integrating a limited workflow with a web application that has no public API.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Hacker News · ksri
Frequently asked questions
How were the APIs documented?
The author navigated the website in Chrome, downloaded network activity as a HAR file, and asked Claude to analyze it and produce OpenAPI JSON.
How did the server obtain authentication?
On connection, it launched a Playwright-controlled browser, checked login status, extracted authentication credentials, cached them locally, and closed the browser.
More details
Remaining steps
- Add Playwright login detection and local credential caching.
About this account
Claude analyzed captured traffic and helped build a local MCP integration; the server automated browser authentication and API calls.
Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=47795216. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: ksri on Hacker News
Source published 16 Apr 2026
Source checked 14 Oct 2026
Tools and topics
Claude Code · Chrome · Python · Playwright · MCP
Software development
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.