Parallel Media Encoder Parallel Media Encoder
The first media encoder MCP built for professional post-production

Media Encoder MCP for Mac. It handles the whole project.

Parallel Media Encoder includes a local Model Context Protocol (MCP) server. Describe the rules once: by resolution, file name, audio channels, frame rate or camera. Your assistant probes every clip, proposes the settings, and fills PME’s hardware queue after you approve the plan. Your media files stay on your Mac.

Download Free Setup guide

MCP server included free. macOS 14.5+, Apple Silicon.

Built for editors, DITs and post teams who already know the codec, colour pipeline and delivery spec they need.

AI assistant · connected to pme MCP
Transcode the whole Harbor shoot on /Volumes/RAID/Harbor. UHD clips to ProRes 422 HQ, HD to 422 LT. Anything with INT in the name is an interview: keep every audio channel and burn in timecode. Everything else gets stereo. Keep source timecode on all of it.
pme_probe_media 148 files 148/148

Found 148 clips from 3 cameras. Here's the plan:

62 3840×2160 · A-cam MXF ProRes 422 HQ · stereo
41 1920×1080 · B-cam MP4 ProRes 422 LT · stereo
18 name contains INT · 8 ch PCM ProRes 422 HQ · all 8 ch · TC burn-in
27 1920×1080 · drone HEVC ProRes 422 LT · stereo
pme_transcode ×4 batches · 148 jobs queued
pme_wait_for_jobs · encoding in parallel 148/148

✓ All 148 done in /Volumes/RAID/Harbor/Transcodes. Heads-up: 2 B-cam clips are 29.97 fps in a 23.976 project. I left them as shot. Want me to import the clips into Resolve?

Illustrative session. Your assistant’s wording will vary.

Works with

Claude DesktopClaude CodeOpenAI CodexGemini CLI

A real session

Ask in your terminal. Watch it in PME.

Claude Code calls PME’s MCP tools, and every job lands in the Encode Queue tagged with the client that sent it. The same slot meters, RAW decode budget and audit results apply as for a job you queued yourself.

Claude Code in Terminal asking PME to make audited ProRes proxies of Blackmagic RAW clips, with the jobs tagged Claude in the Parallel Media Encoder queue

Captured from Claude Code and PME 1.147.02.

One prompt, different settings per clip

A preset treats every file the same. Your AI doesn’t.

PME reports each file’s codec, resolution, frame rate, timecode, colour metadata and audio tracks. The assistant applies your rules to every clip and sends it to the queue with the right settings.

A001C004_250312_R1ZK.MXF

3840×216023.9764 ch
UHD

ProRes 422 HQ

stereo mix

B_0231_INT_Mara.MP4

1920×108023.9768 ch
“INT” in name

ProRes 422 HQ

all 8 ch · TC burn-in

C0412.MP4

1920×108029.972 ch
HD · rate mismatch

ProRes 422 LT

flagged for review

DJI_0087.MP4

3840×2160HEVCno audio
drone

ProRes 422 LT

scaled to 1920×1080

IMG_4471.MOV

1080×1920vertical1 ch
vertical phone

H.264 1080×1920

stereo · social cut

Hours of prep, replaced by a sentence

Real things you can ask. The assistant works out the per-clip settings and PME does the encoding.

Conform a whole project by resolution

“UHD to ProRes 422 HQ, HD to 422 LT, and anything vertical to 1080×1920 H.264 for social. Put it all in /Transcodes.”

Replaces: Sorting hundreds of clips by resolution and building a separate batch for each group.

Route clips by file name

“Anything with INT in the name is an interview: ProRes 422 HQ with timecode burn-in. Clips starting with DJI are drone: 422 LT at 1080p.”

Replaces: Filtering in Finder, dragging each group to the right preset, and hoping nothing got missed.

Decide by audio layout

“If a clip has more than two audio channels, keep every channel for the mix. Phone and GoPro clips get downmixed to stereo.”

Replaces: Opening each file in an inspector to count tracks before choosing settings.

Catch problems before you encode

“Probe the card and tell me any clips whose frame rate or start timecode doesn’t match the rest before you queue anything.”

Replaces: Finding a 29.97 clip in a 23.976 timeline three days into the edit.

Pro

Right LUT on the right clips

“Apply our S-Log3 to Rec.709 LUT only to clips the probe reports as S-Log3. Leave the Rec.709 clips alone.”

Replaces: Checking colour metadata by hand and running two passes with two presets.

Every deliverable from one sentence

“For each final export: an HEVC 10-bit archive master, a 1080p H.264 review copy with timecode burn-in, and a 30-second trim for social.”

Replaces: Three presets, three drags into the queue, and a trim you set by hand for every file.

Chain it with DaVinci Resolve

“Transcode today’s cards with PME, then import the ProRes files into a new Resolve bin named by shoot day.”

Connect Resolve’s MCP server alongside PME and the assistant can take footage from card to timeline-ready in one request.

Where you get the time back

The encoding was already fast. The setup around it wasn’t.

Without MCP

  1. 1 Open clips in an inspector to check resolution, frame rate and audio tracks
  2. 2 Sort files into groups by hand
  3. 3 Build or pick a preset for each group
  4. 4 Drag each group into the queue and set output folders
  5. 5 Come back later to see what failed

With PME + your AI

  1. 1 Describe the rules once, in plain English
  2. 2 The assistant probes every clip and shows you its plan
  3. 3 You approve it, and the jobs go straight into PME’s parallel queue
  4. 4 Watch progress in the app, or ask the assistant for a summary

You stay in control

Built for footage you can’t afford to lose or leak.

Local only

The server and encoder run locally over stdio. PME never sends full media files or audio. Your AI client receives technical metadata and, only when the assistant asks to look at a clip, downscaled still frames.

Originals are safe

PME never overwrites a source. Existing outputs are only replaced if you tell the assistant it’s OK.

Visible in the queue

AI jobs appear in the Encode Queue tagged with the client that submitted them. Pause, reorder or cancel them yourself.

Keeps working after chat closes

With the app closed, a background service keeps encoding. Ask later and the assistant picks the jobs back up.

Built for professional post-production

Generic FFmpeg MCP servers can run commands. PME adds camera RAW, broadcast MXF, source timecode, hardware-aware parallel scheduling and a queue you can control.

Feature Parallel Media Encoder HandBrake EditReady FFmpeg MCP servers
Built-in MCP server Yes No No Third-party wrapper
Runs locally, media files stay on your Mac Yes No AI control No AI control Varies (many are cloud)
Camera RAW (ARRIRAW, R3D, N-RAW, BRAW, Canon RAW) Yes (Pro) No Yes (no AI) No
Hardware ProRes encoding Yes No Yes (no AI) Available in compatible builds
Hardware-aware parallel queue Yes No No Requires manual orchestration
AI jobs visible and cancellable in an app window Yes No AI control No AI control No
Encodes keep running after the chat closes Yes No AI control No AI control No

“-” means not applicable because the app has no MCP server.

Set up in under a minute

1

Install PME

Download the free app, move it to Applications and open it once.

2

Click Install MCP Server

In the app menu, choose Install MCP Server…. PME detects supported MCP clients and configures each one.

3

Ask

Restart your AI app and describe what you want. Try “What’s on this card?” first.

Prefer Terminal, or configuring by hand? See the MCP server guide for every client, the tool reference and troubleshooting.

Frequently asked questions

What makes PME different from a generic video transcoding MCP server?
PME is the first media encoder MCP built for professional post-production. It understands camera RAW, broadcast MXF, source timecode, multichannel audio, LUTs and professional delivery codecs. The MCP server drives the same hardware-aware parallel queue as the app, so jobs remain visible, controllable and grounded in the settings video professionals already use.
Which AI assistants work with it?
PME works directly with Claude Desktop, Claude Code, OpenAI Codex (CLI, desktop app and IDE extension) and Gemini CLI. Any MCP client that can launch a local stdio server should work. ChatGPT can reach local or private MCP servers through Secure MCP Tunnel where that feature is available, but PME’s one-click installer does not currently configure ChatGPT.
Does my footage get uploaded to the AI company?
PME never sends your full media files or audio. The MCP server and encoder run locally on your Mac. Your AI client may send your prompt plus file names, paths and technical metadata such as codec, resolution, frame rate and timecode to its model provider. If the assistant asks to look at a clip, for example to spot sideways shots before choosing a rotation, PME also returns a few downscaled JPEG still frames. Whatever the client receives, frames included, is then handled under that provider’s data policies.
Can the AI overwrite or delete my camera originals?
No. PME never overwrites a source file. If an output file already exists, the encode is refused unless the assistant explicitly sets overwrite, and PME tells the assistant to ask you first. Cancelling a job removes only its partial output.
Is the MCP server free?
Yes. Probing, presets, hardware and license tools are free, and AI-submitted encodes follow the same free-tier rules as the app: H.264, HEVC and ProRes from standard sources, up to 8 queued and 2 at a time. Camera RAW, LUTs, burn-ins, DNx, custom presets and full parallel encoding need Pro ($39, one-time).
Can I watch or stop what the AI is doing?
Yes. While Parallel Media Encoder is open, jobs from your assistant appear in the Encode Queue tagged with the AI app’s name. Pause, reorder or cancel them like any other job. With the app closed, a background service runs them so they keep going after the chat ends.
Does it work with DaVinci Resolve’s MCP server?
Yes. Connect both and the assistant can chain them: transcode with PME, then import the ProRes files into a new Resolve bin. All PME tool names start with pme_ so the two never collide.

Hand your next transcode to your AI

Download Parallel Media Encoder free. The MCP server is included. Unlock camera RAW and full parallel encoding for $39.

Download Free

Requires macOS 14.5 or later. Apple Silicon required.

macOS app · Get the link by email