SkillFoxx

Fitter MCP

A local MCP server where the model writes a JSON config to collect site and API data, validates it and runs it on your machine.

  • 133 ★ on the repository
  • Last commit
  • MIT
MCP server

Extended access

Why this level

  • The agent sends HTTP requests from your address and reads local files through the file source.
  • `fitter_run_url` downloads a config from a link, so use trusted addresses only.
  • HTTP mode without a token is open to anyone who reaches the port.
All reasons and checks

pxyup/fitterAuthor:PxyUp

Install

Manual install

claude mcp add fitter -s user -- "$(pwd)/fitter_mcp"

Download the `fitter_mcp` binary from the releases page first, or build it with `go build -o fitter_mcp ./cmd/mcp`.

Access and checks

How we review

Extended accessThe tool runs code or makes network calls and only reads external data.

  • The agent sends HTTP requests from your address and reads local files through the file source.
  • `fitter_run_url` downloads a config from a link, so use trusted addresses only.
  • HTTP mode without a token is open to anyone who reaches the port.

This is third-party code. Review the repository files before installing.

What it does

Fitter turns a website or API into structured JSON from a declarative config: a source (HTTP request, browser, file, static value) and extraction rules (JSON paths, CSS selectors, XPath). The `fitter_mcp` server gives the agent six tools: the model inspects page structure, writes a config, validates it and runs it. All requests leave your own machine, so no paid scraping API is needed. The config stays a readable file that can be saved and rerun later. The server speaks stdio, with a streamable HTTP mode protected by a token and a Docker image.

Who it is for. Developers and analysts whose agent collects site and API data as JSON and who want to review a config instead of throwaway code.

Good fit when

  • A list page must yield titles and ratings as JSON.
  • An agent picks selectors for an unfamiliar site from the page structure instead of guessing.
  • A config written today becomes a cron job or service tomorrow.
  • A briefing is assembled from several APIs in one run.

Not a fit when

  • The site's rules forbid collecting its data.
  • You need a cloud scraping service without installing a binary.
  • Pages rely on JavaScript and you do not want to install Playwright browsers.

Example request

Take the top five Hacker News stories with titles and scores through fitter and show the JSON.

MCP

Transport
stdio, http
Authentication
not required

Environment variables

FITTER_MCP_AUTH_TOKEN secret
Bearer token for HTTP mode; without it the endpoint is open.
FITTER_MCP_HTTP_ADDR
Listen address for HTTP mode such as `:8080`; empty means stdio.

Limitations

The binary and `.mcpb` bundle ship without browsers: HTTP, static and file sources work at once, while JavaScript pages need Playwright (downloaded on first use with `"install": true` or installed beforehand). The slim Docker image has no browser connectors. HTTP mode without `FITTER_MCP_AUTH_TOKEN` needs no authorization, so bind it to localhost or place it behind a proxy.

How to disable. Remove the connection with `claude mcp remove fitter` and delete the `fitter_mcp` file. In Claude Desktop remove the extension or the `fitter` block under `mcpServers`.

FAQ

Which tools are available?

Six: `fitter_run`, `fitter_run_file`, `fitter_run_url`, `fitter_inspect_url`, `fitter_validate_config` and `fitter_config_reference`. The first three run a config from a string, a file and a URL.

Is a browser required?

Not for HTTP, static and file sources. JavaScript pages need Playwright: install the browsers beforehand or add `"install": true` to the connector.

How do I run a shared server?

Run `fitter_mcp --http :8080` and set `FITTER_MCP_AUTH_TOKEN`. The server answers on `/mcp` and the health probe is on `/healthz`.

Are third-party keys needed?

No. Requests run locally and there is no paid scraping API.

How do I install it in Claude Desktop?

Download `fitter-mcp-<os>-<arch>.mcpb` from the releases page and open the file. Claude Desktop installs the server itself.

Discussion

No questions or reviews yet.

MCP servers
Official

Official Apify MCP server: the agent runs ready-made Apify Store scrapers for social media, maps, search and shops

MCP serverExtended access15.9K

An MCP server that opens pages in headless Chromium through Playwright and hands the agent the main text as Markdown, even from JavaScript-heavy sites

MCP serverExtended access1.1K
Official

Official Firecrawl MCP server: web search, page scraping to Markdown, site crawling and in-page actions

MCP serverExtended access7.8K
Official

Official Oxylabs MCP server: page scraping with JS rendering, Google and Amazon results, an AI crawler

MCP serverExtended access105
Official

Browserless.io MCP server: a cloud browser for the agent with page reads, site crawls, Lighthouse audits and click-driven sessions

MCP serverBroad access19

A local research MCP server: web, Reddit and YouTube search, maps and code exploration for a model in LM Studio or Cursor

MCP serverExtended access531

README in short

One engine runs as an MCP server, a CLI, a Go library and a service with scheduling and notifications. An online demo runs the engine in the browser through WebAssembly.

README badge

Report a problem