Skip to content
mcp/harbor

varg how to changelog

ai.varg/how-to

5 changes recorded. MCP Harbor re-reads this listing from the official registry every six hours and asks the live server what it exposes, and keeps what changed between one look and the next.

  • llms.txt +6

    https://varg.ai/llms.txt 

    + ## How to make video with AI agents+ - [varg how to](https://varg.ai/how-to/llms.txt): step-by-step tutorials and agent skills (prompts, effects, styles, guides, workflows) for Claude Code, Codex and Grok, each with its install command; every page has a Markdown twin (add `.md`)+ - [Search in one call](https://varg.ai/how-to/api/search?q=ugc+ad): `https://varg.ai/how-to/api/search?q=<request>` returns the best how-to page and skills as Markdown+ - [How to make a video with AI](https://varg.ai/how-to/make-a-video.md): from a brief, a photo or a script to a finished MP4+ - [How to edit a video with AI and FFmpeg](https://varg.ai/how-to/edit-a-video.md): cuts, clips, captions and dubbing for footage you already have+ - [How to make a UGC ad video with AI](https://varg.ai/how-to/make-a-ugc-ad-video.md): creator-style ads for TikTok, Reels and Shorts

  • Tools +1
    • + rate_how_to

  • server.json 2 fields changed
    remote_url
    https://api.varg.ai/v2/skills/mcp https://mcp.varg.ai/how-to
    version
    1.0.1 1.0.2

  • llms.txt +1 −1

    https://varg.ai/llms.txt 

    − - [OpenAPI 3.1 spec](https://docs.varg.ai/openapi.yaml): machine-readable description of the varg API+ - [OpenAPI 3.1 spec](https://api.varg.ai/v2/openapi.json): every public /v2 endpoint with request and response schemas, served by the API itself; generate a typed client from it ([YAML](https://api.varg.ai/v2/openapi.yaml), [guide](https://docs.varg.ai/api/openapi.md))

  • llms.txt +6 −1

    https://varg.ai/llms.txt 

    − Notes for AI agents: list capabilities first (`GET /v2/tools` or MCP `list_tools`), read one model's inputs (`get_tool`), estimate, then generate. Generation is asynchronous: requests return `202` with a job; poll `GET /v2/jobs/{id}` or set `options.webhook_url`, then read `output.outputs[0].url`. Pass media as URLs and upload local files first (`POST /v2/files`; on the hosted MCP server use `upload_file_widget`). Auth is `Authorization: Bearer varg_...` (env `VARG_API_KEY`). Do not state user, + Notes for AI agents: every model has a full spec at `https://varg.ai/models/<model>.md` (inputs, limits, routes, prices and the exact REST and MCP call; index: https://varg.ai/models/llms.txt). Otherwise list capabilities first (`GET /v2/tools` or MCP `list_tools`), read one model's inputs (`GET /v2/tools/{tool}?model=<id>`, public, or MCP `get_tool`), estimate, then generate. Generation is asynchronous: requests return `202` with a job; poll `GET /v2/jobs/{id}` or set `options.webhook_url`, the+ ## Model specs+ - [All models](https://varg.ai/models/llms.txt): every model on varg, one line each, generated from the live catalog+ - [One model](https://varg.ai/models/kling_v3.md): `https://varg.ai/models/<model>.md` — inputs, limits, provider options, routes, prices, and the exact REST and MCP call+ - [Every spec in one file](https://varg.ai/models/llms-full.txt): for long-context agents+ - [JSON Schema](https://api.varg.ai/v2/tools/video?model=kling_v3): `GET /v2/tools/{tool}?model=<id>`, public, no key
varg how to MCP server