Broll

io.github.luke-fairbanksPublished by GitHub account luke-fairbanksv0.1.1Updated Oct 24, 2026

Content studio MCP for coding agents: BYO keys, deterministic video rendering, draft-first posting

Overview

AI-generated overview

A local content studio MCP that lets a coding agent generate images and video, render branded carousels and videos from code, and draft posts for social…

What you can do
Broll gives an assistant tools to generate images and video with your own provider keys, import and probe media assets, and render deterministic videos and branded carousels from declarative plans using ffmpeg and sharp. It can extract a frame for visual QA, create validated post drafts in a local outbox, list drafts, and publish them. Publishing requires an explicit confirm flag and validates per-platform rules before anything leaves the machine.
When to use it
Use it when a coding agent needs to produce marketing or social media assets and posts rather than just plan them, and when you want reproducible, code-driven rendering instead of a credit-metered scheduler. It suits developers who already have their own OpenAI or Gemini keys and want draft-first control over what gets posted.
Requirements
Runs locally as an npm package (broll-mcp) over stdio, requiring Node 20+ and ffmpeg installed on the machine. Optional environment variables enable features: OPENAI_API_KEY for image generation, GEMINI_API_KEY for images and video, BLUESKY_IDENTIFIER and BLUESKY_APP_PASSWORD for Bluesky posting, MASTODON_ACCESS_TOKEN for Mastodon, and X_API_KEY, X_API_SECRET, X_ACCESS_TOKEN, X_ACCESS_TOKEN_SECRET for X. A mock provider works without any keys.
Before you install
Live posting sends content to third-party platforms and requires the platform credentials listed above; X posting is noted as requiring funded API credits. Generated media is sent to the vendor of whichever key you supply. Publishing is gated behind confirm: true and a draft file on disk, but an agent with the tool could still post if it sets that flag, so review drafts before granting go-ahead.

Installation

In SourceWeft

  1. Open Broll in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

broll

The content studio MCP for coding agents. broll gives Claude Code, Codex, and any MCP client real hands for content work: generate media with your own API keys, render videos and carousels deterministically with code, and publish through a draft-first outbox.

The model plans. Code renders. Nothing posts without confirmation.

Why

Every developer running a coding agent hits the same wall: the agent can write the marketing plan, but it can't make the carousel, cut the video, or post it. Existing tools are credit-metered schedulers built for social media managers. broll is built for developers:

  • BYO keys, no markup. Image/video generation uses your OPENAI_API_KEY / GEMINI_API_KEY directly. broll never proxies your inference or resells credits. A built-in mock provider keeps every workflow runnable before you add any keys.
  • Deterministic rendering. Videos and slides are compiled from declarative plans into exact ffmpeg/sharp invocations. Same plan + same inputs = same output. When AI output drifts, the fix is code — so layout, fonts, captions, and branding live in code.
  • Draft-first publishing. Posts are reviewable JSON files in an outbox. publish_post requires confirm: true, validates per-platform rules (char limits, media counts, file sizes) before anything leaves the machine, and reports per-platform results. The export platform always works: it writes a ready-to-post bundle.

Install

Requires Node 20+ and ffmpeg (brew install ffmpeg).

bash
cd brollnpm install && npm run build

Register with Claude Code (or use the checked-in .mcp.json when working inside this repo):

bash
claude mcp add broll -- node /path/to/broll/dist/index.js

Optional environment:

VariablePurpose
OPENAI_API_KEYgpt-image-1 image generation
GEMINI_API_KEYImagen images + Veo video generation
BLUESKY_IDENTIFIER / BLUESKY_APP_PASSWORDlive Bluesky posting (use an app password)
MASTODON_ACCESS_TOKEN (+ optional MASTODON_INSTANCE)live Mastodon posting — no app review, any instance
X_API_KEY / X_API_SECRET / X_ACCESS_TOKEN / X_ACCESS_TOKEN_SECRETlive X posting (free API tier works)
BROLL_HOMEworkspace location (default ~/.broll)
BROLL_FFMPEG / BROLL_FFPROBEexplicit binary paths

Brand kit: drop a broll.config.json next to where the server runs (see this repo's for an example) — name, handle, colors, font, logo. Every render picks it up automatically.

Tools

ToolWhat it does
broll_statusWorkspace, ffmpeg, provider + platform readiness. Call first.
generate_image / generate_videoBYO-key generation → workspace assets (mock fallback when keyless)
import_asset / list_assets / probe_assetBring in and inspect media
render_videoDeclarative RenderPlan → mp4: clips (video/image/color), trims, cover/contain fits, burned-in titles + timed captions, music bed, 9:16 / 1:1 / 4:5 / 16:9
render_carouselBranded slide sets (the Instagram/LinkedIn format): kicker, headline, body, accent bar, page numbers, watermark — layout is 100% code
extract_framePull a PNG frame so the agent can visually QA its own render
create_post_draftText + media + platforms → validated, reviewable draft in the outbox
list_drafts / publish_postInspect the outbox; publish with explicit confirm: true

Try it

bash
npm run smoke

drives the real server through a real MCP client: generates a background, renders a 3-slide carousel and a 9:16 teaser video with captions, extracts a QA frame, then drafts and "publishes" an export bundle — all into ./.broll/.

broll cloud (waitlist)

The local server you're looking at stays MIT and free. broll cloud adds the parts a local server can't do well: the app-review-walled platforms (Instagram, LinkedIn, TikTok) through broll's own approved credentials, schedules that run while your laptop is closed, and a swipe-to-approve inbox on your phone. ~$29/mo when it ships.

Join the waitlist → 👍 issue #1 — subscribers get the launch note; comments set the platform build order.

Safety model

  1. Nothing is published without a draft file on disk first.
  2. publish_post hard-requires confirm: true — agents are instructed to obtain the user's explicit go-ahead.
  3. Constraint violations block publishing; they never auto-truncate your text.
  4. Keys are read from your environment and sent only to their own vendor's API.

Status & roadmap

Early but real: 96 tests including real-ffmpeg integration renders and a full MCP round trip.

  • Bluesky video upload
  • X chunked video upload
  • LinkedIn adapter (needs app review)
  • Mastodon adapter
  • YouTube Shorts via user OAuth
  • Ken Burns / crossfade transitions
  • Word-level caption timing from transcripts
  • npx broll-mcp distribution

Development

bash
npm test          # unit + integration (real ffmpeg)npm run typechecknpm run smoke     # end-to-end artifact check

MIT.

Source: README.md at commit a5b8a67

Tools

0
SourceWeft has not started this package in its sandbox yet, so its tools are not known.

Version history

1
  1. v0.1.1LatestSep 16, 2026

Score

50% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts

    Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.

  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts