#99th of 12 in MCP servers
pomerado
Generates MCP servers for websites from a plain-language task description
Documentation ↗
(opens in a new tab)Website ↗
(opens in a new tab)Repository on GitHub ↗
(opens in a new tab)
- Stars
- 1.6k
- License
- MIT
- Last commit
- Nov 2026
- Last release
- Nov 2026
- Language
- TypeScript
2 on 6 Oct 20261,590 on 10 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Pomerado drives a real Chromium browser to explore a site, complete an example task, and generalize it into a deterministic integration saved as TypeScript source. Each integration runs as its own local MCP stdio server and makes no model call at run time. Minting uses OpenAI models and a review step called Guardian that checks browser actions and the final source.
Who it is for: Engineers who want agents to use websites that have no API
Strengths
- Needs only a text task description; no HAR file or recording required
- Minted integrations run with no model request; only minting needs an API key
- Works on sites behind login and two-factor authentication without saving credentials
- MIT core includes minter, Guardian and local host; no account required
Weaknesses
- Minting requires an OpenAI API key; no other provider is listed for building
- Supports macOS and Linux only, and Node 24.21 up to below 25
- Each mint is limited to 20 minutes of active work
- Hosting, saved logins and automatic repair when sites change are Cloud-only
What it needs
- no GPU
- Needs Node 24.21+ (<25), Chromium via Playwright, OpenAI API key, MCP client
- Models: OpenAI (minting), any model for the calling agent
Quick answers
Is pomerado open source?
Yes. Its code is on GitHub under the MIT license.
Does pomerado need a GPU?
No, it runs without one.
Also in MCP servers
See all 12| Rank | Project | Score |
|---|---|---|
| 1 | serversReference MCP servers for files, git, fetch, memory and time | 94 out of 100 |
| 2 | apify-mcp-serverOfficial MCP server exposing Apify Store Actors as tools for web scraping | 86 out of 100 |
| 3 | xiaohongshu-mcpMCP server that lets AI clients post, search and comment on Xiaohongshu | 85 out of 100 |
| #4 | ha-mcpMCP server that lets AI assistants configure and control Home Assistant | 80 out of 100 |
| #5 | mcp-atlassianMCP server exposing Jira and Confluence tools to AI assistants | 69 out of 100 |