What it does

Rendobar is a serverless media processing API. You POST a job with your inputs. Rendobar runs it on isolated cloud workers. The output comes back by webhook or polling. Rendobar handles provisioning, the media toolchain, sandbox isolation, scaling and cleanup. Each job runs in an ephemeral container with protocol whitelisting and an automatic timeout. Rendobar handles input download, execution, output upload and delivery. Nine job types are live. They include ffprobe metadata reads, compress.target for shrinking a file to a target size or quality, and image generation, editing and upscaling. Jobs can run on CPU or GPU compute, sized to each job. Styled QR codes and managed watermarks are coming soon.

Capabilities

The FFmpeg API runs FFmpeg 9.0 from one API call. You send the command you would run locally and get back every file it writes.
Compose renders a finished MP4 from one JSON request, with clips, titles, captions and music timed for you. It supports up to 4K at 60 fps.
AI Captions burns word-level animated captions into a video, auto-transcribed, styled and translated in one call. Static subtitles can also come from an SRT or VTT file.
Rendobar ships two Model Context Protocol servers. The hosted one at api.rendobar.com/mcp exposes 11 tools over OAuth. The @rendobar/mcp npm package runs on your machine so agents in Claude Desktop, Cursor and Zed can upload local files into a job.
Rendobar sends a signed webhook when a job finishes, or streams every update live over SSE.

When to use it

You want to add video processing to your app without running servers.
You need to burn animated word-level captions into a video.
You want agents in Claude, Cursor or another MCP client to call media jobs as native tools.