capturekit

capturekit

Page content as clean Markdown for the built-in web_extract tool, through the CaptureKit Content API. Extract only; pair it with a search backend for web_search. Disclosure — sends each URL to api.capturekit.dev with your CAPTUREKIT_API_KEY, and calls can use CaptureKit credits. Reads the key only through Hermes's secret scope. Registers no tools, no hooks or middleware, writes no files, starts no processes, has no telemetry or self-updater, and uses only the Python standard library.

capturekitdevWebv1.0.0Updated 22 minutes ago

About

CaptureKit for Hermes Agent

Page content as clean Markdown for the built-in web_extract tool, through the CaptureKit Content API.

Install

hermes plugins install capturekit

Hermes asks for CAPTUREKIT_API_KEY during the install. Create a key in the CaptureKit dashboard.

Then select it in hermes tools, or in ~/.hermes/config.yaml:

web:
  extract_backend: capturekit

It is extract only, so keep a search backend for web_search.

Settings

| Setting | Default | Meaning |
|---|---|---|
| main_content_only | true | Keep the main content of each page and drop navigation, footers and sidebars. |

What it does on your machine

Network: calls https://api.capturekit.dev only, with your CAPTUREKIT_API_KEY. Calls can use credits on your CaptureKit account (pricing).
Credentials: reads the key only through Hermes's secret scope. It stores nothing.
Nothing else: no files written, no processes started, no telemetry, no self-update. Standard library only, no dependencies.

Development

python -m pytest

The tests mock the network and run without Hermes installed.

Contact

Source: https://github.com/capturekitdev/hermes-plugin-capturekit
CaptureKit: https://www.capturekit.dev
Issues: https://github.com/capturekitdev/hermes-plugin-capturekit/issues

License

MIT

README from the reviewed commit 4b74a56; it updates when the author re-pins.