Tell your computer what to do. In plain English. Agentino plans it, picks the right tools, and runs it end-to-end — on your own machine.
Agentino is a CLI-first agentic OS. You describe a task in a single sentence — "analyse this video and trim the silences", "turn this script into a captioned explainer", "back up my databases every night" — and Agentino routes it through local skills, chains them into pipelines, and hands you the result. No hosted control plane, no SaaS subscription for the runtime, no vendor lock-in on the code that actually runs.
- Local-first. Your files, your API keys, your machine. Nothing runs in someone else's cloud unless you explicitly call a cloud skill.
- CLI-first. Born in the terminal.
agentino run "…"is the primary interface; a REST API, live dashboard, and IDE integrations layer on top. - Marketplace-backed. Every skill is a public GitHub repo.
agentino marketplace install <owner>/<repo>and you have a new verb. Share yours the same way. - Composable. Skills chain into pipelines. Pipelines chain into pipelines. The LLM planner is just one kind of scheduler — declarative YAML works the same way.
- Batteries included. 14 skills and 3 pipelines out of the box for video, plus TTS, image generation, LLM routing across OpenAI / Anthropic / Gemini / Ollama / OpenRouter / LiteLLM.
# Python 3.12+
pip install agentino
# Ask for something
agentino run "hello world"
# Or browse + install a ready-made skill
agentino marketplace search video
agentino marketplace install agentino-os/agentino-skill-video-brief
# Run the full generative pipeline (requires an OpenAI or OpenRouter key)
agentino marketplace install agentino-os/agentino-skill-video-image-from-prompt
agentino marketplace install agentino-os/agentino-skill-video-storyboard-from-script
agentino marketplace install agentino-os/agentino-skill-video-voiceover
agentino marketplace install agentino-os/agentino-skill-video-slideshow
agentino marketplace install agentino-os/agentino-skill-video-captions-burn
curl -fsSLo pipelines/script-to-video.yaml \
https://raw.githubusercontent.com/agentino-os/agentino-pipeline-script-to-video/v1.0.0/pipeline.yaml
agentino pipeline run script-to-video \
-i script="Agentino runs on your machine. Local-first. Automate anything you can describe." \
-i shot_count=4 \
-i output_path=/tmp/pitch.mp4Full guide: agentino-os/agentino.
| Repo | What it is |
|---|---|
agentino |
The runtime: CLI, planner, skill registry, marketplace client, HTTP API. |
marketplace |
Canonical skill registry. agentino marketplace search reads from here. |
| Skill / pipeline | What it does |
|---|---|
video-brief |
LLM-readable brief of a video — metadata, silence gaps, scene changes, thumbnails. |
video-cut-silences |
Remove dead air with inaudible crossfades at every cut. |
video-concat |
Glue N clips into one MP4, normalising res / fps. |
video-intro-outro |
Prepend branded intro, append outro, with smooth crossfade. |
video-chapters |
YouTube / QuickTime chapter markers via ffmetadata (no re-encode). |
video-speed |
Time-warp 0.25× to 4× with pitch-preserving audio. |
video-auto-trim |
Pipeline: brief + cut in one command. |
| Skill / pipeline | What it does |
|---|---|
video-voiceover |
Text → narrated WAV + word-timed SRT (edge-tts / say / piper). |
video-slideshow |
Images + optional audio → Ken Burns slideshow MP4. |
video-captions-burn |
Burn SRT / VTT subtitles into the video stream. |
video-bgmusic |
Music bed with automatic voice-over ducking. |
video-lowerthirds |
Timed text bands (9 positions), configurable font / colour. |
video-pip |
Picture-in-picture composition (webcam overlay, multi-cam). |
slides-to-video |
Pipeline: script + images → captioned narrated MP4 (pure local). |
| Skill / pipeline | What it does |
|---|---|
video-image-from-prompt |
Text → PNG via OpenAI Images or OpenRouter (Gemini Flash Image, Flux…). |
video-storyboard-from-script |
LLM plans N shots, image model renders each. |
script-to-video |
Pipeline: script → AI storyboard + narration → captioned MP4. |
More themes on the roadmap: audio/, system/, git/, text/.
A skill is a single skill.yaml (plus an optional Python / bash / LLM executor block) shipped as a public GitHub repo. The marketplace client clones it, validates it against a security scanner, and installs it locally.
Read the 10-minute walkthrough: Marketplace Skill Author Guide.
- Discussions → agentino-os/agentino/discussions
- Bugs / feature requests → agentino-os/agentino/issues
- Docs → the in-repo guide (a public site is on the way)
- Contributing → CONTRIBUTING.md
MIT across the whole ecosystem. Use it, fork it, ship it.
Agentino is built in the open by @dagoSte and contributors. If you want to help shape the roadmap — picking the next theme after video/, proposing a skill you would install today if it existed, or fixing a rough edge — open an issue or discussion. Welcome in.