Coding agents

​

Install

• Cursor
• Claude Code
• VS Code
• Codex
• Gemini CLI

Add to ~/.cursor/mcp.json (or project-local .cursor/mcp.json):

{
  "mcpServers": {
    "rendi": {
      "url": "https://www.rendi.dev/docs/mcp"
    }
  }
}

Or click this button:

claude mcp add --transport http rendi https://www.rendi.dev/docs/mcp

code --add-mcp '{"name":"rendi","type":"http","url":"https://www.rendi.dev/docs/mcp"}'

Or add to .vscode/mcp.json:

{
  "servers": {
    "rendi": {
      "type": "http",
      "url": "https://www.rendi.dev/docs/mcp"
    }
  }
}

Add to ~/.codex/config.toml:

[mcp_servers.rendi]
url = "https://www.rendi.dev/docs/mcp"

gemini mcp add rendi --transport http https://www.rendi.dev/docs/mcp

​

Claude Code: FFmpeg plugin

/plugin marketplace add rendi-api/ffmpeg-llm
/plugin install ffmpeg-llm@rendi-official

​

AGENTS.md snippet

This project uses Rendi for running FFmpeg commands. Rendi evolves, always check the latest documentation at www.rendi.dev/docs. For FFmpeg command patterns: https://raw.githubusercontent.com/rendi-api/ffmpeg-cheatsheet/refs/heads/main/README.md — and for an FFmpeg authoring skill: https://raw.githubusercontent.com/rendi-api/ffmpeg-cheatsheet/refs/heads/main/skills/ffmpeg/SKILL.md

​

Other ways to access Rendi docs

• OpenAPI spec — canonical request and response schema for every endpoint — https://www.rendi.dev/docs/api-reference/openapi.json
• llms.txt (page index) — https://www.rendi.dev/docs/llms.txt
• llms-full.txt (full docs concatenated) — https://www.rendi.dev/docs/llms-full.txt
• skill.md (condensed summary of how Rendi works, auto-generated) — https://www.rendi.dev/docs/skill.md
• Portable FFmpeg skill — drop-in SKILL.md for any skill-aware agent — https://raw.githubusercontent.com/rendi-api/ffmpeg-cheatsheet/refs/heads/main/skills/ffmpeg/SKILL.md
• Raw markdown of any page — append .md to any docs URL (e.g., /docs/coding-agents.md)

​

Best practices for agents using Rendi

1. Use in_* / out_* key names — keys in input_files (e.g. in_1, in_video) and output_files (e.g. out_1, out_gif) must use these prefixes. The same keys are referenced as {{in_1}} / {{out_1}} placeholders inside ffmpeg_command.
2. Pass any publicly accessible HTTP/HTTPS URL as an input — Google Drive share links, Dropbox, public or signed S3/GCS, and CDN URLs all work as-is. If it loads in an incognito browser, Rendi can fetch it.
3. Set max_command_run_seconds to match the command’s expected runtime — it’s a safety cap that aborts runaway FFmpeg jobs (defaults to 5 minutes). Pick a value that covers legitimate cases but still catches hangs.
4. Rendi stores your output files for you — use the storage_url returned in the response directly, and use store-file / get-file for persisting intermediate assets. You don’t need to store output files in another S3/GCS unless your app specifically requires it.
5. Use run-chained-ffmpeg-commands for multi-step pipelines — single round-trip, lower latency than chaining separate calls.
6. Use webhooks for jobs over ~30 seconds; poll for shorter ones.
7. Use dynamic-files when the same command runs across many inputs.
8. Authentication is the X-API-KEY header — keep keys server-side, never in browser code. See API introduction.
9. For FFmpeg commands not covered by an example page, fetch the Rendi cheatsheet: https://raw.githubusercontent.com/rendi-api/ffmpeg-cheatsheet/refs/heads/main/README.md — it lists canonical commands for common operations including audio mixing, crossfades, jump cuts, social-media cropping, frame-rate changes, and storyboards.

​

Suggested prompts

• “Add a Rendi-based thumbnail extractor to my Express route handler.”
• “Build a video compression pipeline using Rendi with webhook callbacks.”
• “Extract audio from this MP4 using Rendi and save as MP3.”
• “Use Rendi to burn SRT subtitles into a video, then compress the result to under 5MB.”
• “Generate 10 thumbnails evenly spaced across a video using Rendi, in one chained call.”