The higgsfield supercomputer github search leads to a repo that wraps Higgsfield’s rendering stack for agent builds. Developers use it to script video generation, then need a publishing layer. Creator OS is that layer: one post can carry the finished clip to Instagram, TikTok, YouTube, X, Facebook, LinkedIn, Threads, Reddit, Discord and Slack, with the API, CLI and hosted MCP server included in every plan.
higgsfield supercomputer github repo
|
v
+-----------------------+
| render clips by code | agent or script drives it
+-----------------------+
|
v
+-----------------------+ +--------------------+
| POST /v1/media | ---> | med_ id, reusable |
+-----------------------+ +--------------------+
|
v
+-----------------------+
| POST /v1/posts | ---> Instagram, TikTok, YouTube,
| platforms: [...] | X, Facebook, LinkedIn,
+-----------------------+ Threads, Reddit, Discord
|
v
analytics: watch time, skip rate, best time to post
What Is the Higgsfield Supercomputer GitHub Repo?
The repo is the code side of Higgsfield’s Supercomputer, packaged so a developer or an AI agent can drive renders from a script instead of a browser tab.
Read it as a build tool, not a scheduler. You get a way to call the generation stack from code, which matters when you produce ten or fifty clips in a batch and never want to click a download button. What the repo does not do is carry that output anywhere. There is no publishing step, no per-platform caption handling, no scheduled slot, no comment inbox. That gap is the whole reason a publishing layer exists next to it.
If you want the publishing half described in one pass, read the Creation OS explainer first, then come back here for the build loop.
How to Find and Use higgsfield.ai/supercomputer
The product page at higgsfield.ai/supercomputer is the front door; the repo is where you copy the calls that page describes.
- Open the Supercomputer page and read what the current build supports before you plan anything around it.
- Check the repo’s README and examples for the call pattern, keys and limits. Treat the README as the only authority; anything else is guesswork.
- Run one small render end to end before you write batch logic around it.
- Store the resulting file or URL where your publish step can reach it. Creator OS accepts a public URL or an uploaded media id, so a hosted file is the simplest handoff.
- Write the publish call last, so a render failure never leaves a half-finished post in the queue.
One habit saves real time: keep the render step and the publish step in separate functions. When Higgsfield ships a change, you fix one function. Your captions, schedules and platform rules stay untouched.
Is Higgsfield Supercomputer Free?
Whether Supercomputer costs money is a question for its own pricing page and repo, and both can change between releases. Check current, first-party terms before you budget for volume. Two things hold firm regardless of how that pricing lands.
- A free render tier, if one exists, still owes you nothing on distribution. Rendering and publishing are separate costs from separate vendors.
- Your publishing cost with Creator OS is published and flat. Creator Premium is $14.99/month or $59.99/year, roughly $5/month, and covers one set of socials: 11 platforms plus Skool, one account on each, plus the iOS app.
Do the arithmetic as cost per published clip rather than cost per render, because a clip that never leaves a folder has zero value. New subscribers start for $0 with a free trial and can cancel anytime.
Higgsfield Supercomputer API and MCP for Agent Builds
Agent builds get reliable when both ends of the loop speak a protocol the model already understands.
On the Creator OS side, the hosted MCP server at https://mcp.creatoros.ca/mcp needs nothing installed. In Claude, open Settings > Connectors > Add custom connector, paste the URL, sign in, pick the workspace, then choose read only or read & write. ChatGPT takes the same URL as a connector in developer mode. Claude Code, Cursor, Windsurf and VS Code take an mcpServers entry with type http and a Bearer key. The server exposes about 160 tools, including create_post, upload_media_from_url, get_post_analytics, get_best_time_to_post, reply_to_comment and check_caption_length.
Read-only connections only see read tools, and the API refuses writes from them. Destructive tools such as delete post or disconnect account are marked so the AI app asks before acting.
For a code-first loop, use the CLI after one setup step:
npx @creatoros/cli@latest init
creatoros posts:create --text "clip caption" \
--platforms instagram,tiktok,youtube \
--media https://cdn.example.com/clip.mp4 \
--scheduledAt 2026-03-04T18:00:00Z --timezone America/Toronto
The matching REST call is a single POST to /v1/posts with content, platforms, media, schedule_at and timezone. Get the base URL and field list from the docs. Keys look like cos_live_ and are pinned to one workspace, so a key issued for one client cannot post to another. A post that names an unconnected network comes back with missing_platforms and the rest still go out.
Creator OS does not ship its own model. It works with any MCP-capable AI app or agent.
How the GitHub Higgsfield Repo Fits With Motionmaxxing GitHub
The two repos answer different questions, which is why they pair cleanly instead of competing.
Motionmaxxing on GitHub is about motion design: animation, timing, the look of the thing. The Higgsfield repo is about generating footage. Put together, you get a rendered clip that also moves well. Neither one posts.
There is a third open tool worth naming for this loop. OPEN VIDEO EDIT, at github.com/kevinbadi/open-edits, is a set of open-source Claude Code skills that turn a raw talking-head clip into an animated short with captions, logos, b-roll cards and counters, rendered locally with Python and ffmpeg. Skills include split-animated-talking-head, fireship-style-short and reverse-engineer.
The full chain reads like this: generate with the Higgsfield stack, style with Motionmaxxing or OPEN VIDEO EDIT, publish with Creator OS, then read the numbers back. For background, the Motionmaxxing write-up and the Higgsfield repo guide cover their own halves.
| Step | Tool | Output |
|---|---|---|
| Generate footage | Higgsfield Supercomputer repo | Video file or URL |
| Style and animate | Motionmaxxing, OPEN VIDEO EDIT | Edited vertical short |
| Host media | POST /v1/media or creatoros media:upload | med_ id reusable per platform |
| Publish | POST /v1/posts or posts:create | Live posts, one call |
| Measure | get_post_analytics, analytics:best-time | Skip rate, watch time, next slot |
How Creators Turn Higgsfield Renders Into Published Posts
A render becomes useful when an agent can publish it, read the result and change the next render without a human in the middle.
Take one vertical clip. Upload it once and reuse the id:
creatoros media:upload ./clip.mp4
creatoros posts:create --text "three takes on one hook" \
--accounts acc_instagram,acc_tiktok,acc_youtube \
--media med_9f3c \
--draft
creatoros validate:post-length --text "three takes on one hook"
The validator catches a caption that runs past a platform’s limit before you schedule, which is cheaper than a failed post at 6pm. Then let the post go live, and read it back. Instagram Reel analytics expose reelsSkipRate, the share of views that swiped away in the first three seconds, plus average watch time and total view time in milliseconds. Those three numbers tell an agent whether the hook held. A prompt can say: if skip rate is high, shorten the opening and reschedule the next render for the best slot returned by get_best_time_to_post.
Creator OS supplies the posting and the numbers. Your prompt or workflow decides what changes. That boundary keeps the loop predictable and auditable.
The same loop works for a batch. Write a calendar file and dry-run it first:
creatoros posts:bulk-upload --file calendar.json --dryRun
Comments come back into the same inbox. Instagram and Facebook support comment-to-DM keyword funnels, so a reply under a generated Reel can trigger a DM automatically, and the comments-to-DM guide shows the setup. If you build with an agent harness, Social Agents at github.com/kevinbadi/social-agents runs your socials on Creator OS and starts with npm start creatoros social-agents; it interviews you about your brand, then posts, replies and reports. To watch the build in motion, see the launch video and the walkthroughs on the KevBuildsApps YouTube channel.
FAQ
Is Higgsfield Supercomputer free?
Check the current first-party pricing page and repo before you plan volume, since that answer can change between releases. What does not change: rendering and publishing are separate costs, and Creator Premium is $14.99/month or $59.99/year for 11 platforms plus Skool. New subscribers start for $0 with a free trial and can cancel anytime.
How do I access the Higgsfield Supercomputer API?
Follow the repo’s README and the Supercomputer product page for keys, call shape and limits; those two sources are the authority. Creator OS is a separate API for the publishing side, authenticated with a cos_live_ key pinned to one workspace. Base URL and request fields live at https://www.creatoros.ca/docs.
What is the Higgsfield Supercomputer MCP used for?
The purpose of any MCP server is to let an AI app call a tool directly instead of asking you to copy and paste. Creator OS runs a hosted MCP server at https://mcp.creatoros.ca/mcp that exposes about 160 publishing, analytics, inbox, automation and ads tools to Claude, ChatGPT, Claude Code, Cursor, Windsurf and VS Code. Setup steps are at https://www.creatoros.ca/docs/mcp.
What is the Motionmaxxing GitHub repo and how does it compare?
Motionmaxxing is about motion and animation craft, while the Higgsfield repo is about generating footage. They stack rather than compete: generate, then style, then publish. Creator OS handles only the last step, plus the analytics that tell you what to make next.
Can I use the higgsfield repo to auto-post generated videos?
No, the repo covers rendering; distribution is a separate call. Send the finished file to POST /v1/media once, then POST /v1/posts with the platforms you want, and Creator OS applies each network’s rules such as Reels, Shorts and TikTok consent flags. A read and write MCP connection or a CLI command does the same job from inside the agent.
Get started
Creator OS plans start at $14.99/month for 11 platforms plus Skool, with the API, the MCP server and the CLI included. Agency plans add one client workspace per set of socials and a scoped key for each. You can sign up here and connect your first accounts in a few minutes.
For the wider build pattern, read the GitHub motion maxxing guide and how to schedule faceless videos, then watch more walkthroughs on the KevBuildsApps channel.
Get started for $0
Free trial for new subscribers. Cancel anytime.
