podcli MCP server
Commit Nov 3, 2026Turns long podcast episodes into short vertical clips with face tracking and burned-in captions through an AI agent.
Also known as nmbrthirteen/podcli
- Stars
- 89
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- —
- Last commit
- Nov 3, 2026
- License
- AGPL-3.0
What is podcli MCP?
podcli turns long podcast episodes into short vertical clips with face tracking and burned-in captions through an AI agent.
Summary generated from the project's README
About podcli MCP server
Transforming long podcast episodes into short vertical clips with face tracking and burned-in captions is what podcli enables for content creators. This open-source tool allows users to automate the tedious process of finding highlights, cropping video to follow speakers, and adding various caption styles like karaoke or branded layouts. It runs locally on macOS, Linux, and Windows, ensuring that transcription and rendering happen on your own machine to keep data private.
Developers can interact with the system through a command line interface, a web-based studio at localhost:3847, or by connecting it to an AI agent via 32 MCP tools. While transcription and rendering are local, clip scoring can utilize an Anthropic API key or a Claude Code or Codex login. For advanced speaker tracking, the system can integrate with AssemblyAI. The server supports complex workflows including multicam syncing, loudness-normalized audio, and exporting timelines to professional software like Premiere or Final Cut.
What you can do with it
- 1Automated clip generationAn agent can transcribe an episode and automatically render short clips with captions.
- 2Multicam episode editingSynchronize multiple camera and microphone files to create a polished full-length episode.
- 3AI-driven highlight discoveryUse AI to score and identify the most engaging moments from a transcript.
- 4Professional video handoffExport your edits as specialized files to finish manual work in DaVinci Resolve.
Written from the project's README. Every use case is checked against a quote from it.
Automated checks
Ran Nov 14, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachablenmbrthirteen/podcli is public and reachablegithub.com/nmbrthirteen/podcli ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)AGPL-3.0spdx.org/licenses/AGPL-3.0.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-11-03github.com/nmbrthirteen/podcli ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHubpodcli process episode.mp4
That one command transcribes the episode, picks the moments worth clipping, crops to the visible face, and burns the captions in. Transcription and rendering run on your machine. Clip scoring uses Claude Code or Codex login, or an Anthropic API key; without those it falls back to heuristics.
Install
No prerequisites. The installer fetches a self-contained binary, and the first run provisions Python, Node, FFmpeg, whisper.cpp, and the models it needs into a managed folder.
macOS and Linux
curl -fsSL https://podcli.com/install.sh | sh
Windows (PowerShell)
irm https://podcli.com/install.ps1 | iex
Runs on macOS (Apple Silicon), Linux (x64 and arm64), and Windows (x64). Intel Macs aren't supported.
Quick start
podcli # interactive menu, opens the web studio
podcli process episode.mp4 # transcribe, pick moments, render clips
podcli multicam ~/ep12 # sync every camera and mic, cut to the speaker, render the episode
Clips land in podcli-clips/ in the directory you ran it from, so each show keeps its own renders. Everything else (knowledge, presets, assets, clip history, cache) lives in one managed folder that follows you between directories. Set PODCLI_OUTPUT to render somewhere fixed instead.
What you get
Clips
- 9:16, 16:9, or 1:1, with captions sized for each canvas
- Face tracking that follows the visible face, split-screen layouts included. Following the active speaker needs AssemblyAI, or a source install with openai-whisper and
HF_TOKEN - Multi-segment cuts that drop filler, long pauses, and tangents
- Four caption styles: branded, hormozi, karaoke, subtle
- Logos, intros, outros, and background music from a reusable asset library
- Loudness-normalized audio and hardware encoding on VideoToolbox and NVENC, with a CPU fallback
Editing the full episode
- Multicam edit: point at one folder of camera and mic files, confirm who each file belongs to, and podcli syncs them by audio and corrects clock drift between devices
- Automatic camera cuts that follow whoever is talking, with wide shots for crosstalk and a guest's answer held on the guest
- Remote shows too: one file per person becomes a split screen, and a single gallery recording is split into a camera per tile. Questions stay on the split and the guest goes full frame on long answers
- Render the episode as an MP4 with one WAV per person. Studio shoots can also go to Premiere or Final Cut as a timeline that points at your original files
- Same edit from the studio (Multicam edit), the CLI (
podcli multicam <folder>), or an agent (manage_multicam). Re-running reuses the sync and the cut, so only the step you changed runs again - With podcli Pro, steer the cut in the cloud editor:
podcli multicam <folder> --clouduploads per-camera proxies, a mic mix (about 550 MB for a two-hour, three-camera episode), and the transcript. Your original camera files stay on your computer, and--pullrenders the cut you made there
Finding the moments
- Whisper transcription (speaker diarization on AssemblyAI, or openai-whisper from a source install or
podcli setup --speakers), or bring your own transcript as.txt,.srt, or.vtt - AssemblyAI as an alternative engine, and yt-dlp to pull an episode straight from a URL
- AI scoring against your knowledge base, checked against your episode database so it stops resuggesting moments you already published
- Audio energy and laughter detection to build highlight reels
The studio at localhost:3847
- Library, episode workspace, multicam edit, per-clip detail, highlights, thumbnails, content, analytics, assets, knowledge, config, integrations, and MCP setup
⌘Kcommand palette across pages, clips, and assets- Titles, descriptions, tags, and hashtags, with any section regenerated on your own guidance
- Thumbnail studio for 16:9 and 9:16, with frame and text options
- Transcript corrections that carry through to every render
Shipping it
- 32 MCP tools, so an agent can transcribe, score, and render through conversation
- YouTube analytics to see which clips landed
- DaVinci Resolve export as FCPXML when you want to finish by hand
- Presets, clip history with duplicate detection, and a transcript cache
Why podcli
If you are weighing podcli against the cloud clippers, this is the difference:
- Runs locally. Transcription and rendering happen on your machine by default, so episodes stay there. Data leaves your machine only through these destinations:
- AssemblyAI, if you pick that engine: the episode audio.
- AI scoring and titles: the transcript goes to podcli Pro when you are signed in, to Anthropic when you set
ANTHROPIC_API_KEY, and to the provider behind your Claude Code or Codex login. - podcli Pro, when signed in: knowledge and asset sync, plus rendered clips and transcript slices if you turn on
PODCLI_CLOUD_SYNC=1(off by default). - YouTube, if you link it: read-only analytics requests.
- podcli multicam cloud, if you run
podcli multicam --cloud: per-camera proxies, a mic mix, and the transcript.
- Free and open source under AGPL-3.0. Exports are unlimited, full quality, and watermark-free.
- Agent-native. 32 MCP tools let Claude Code or Codex drive the whole flow, from transcription to rendered clips and publish-ready titles.
- A knowledge base keeps titles, captions, and descriptions in your show's voice, and stops the engine from resuggesting moments you already published.
- DaVinci Resolve handoff. Export any clip as FCPXML when you want to finish the edit yourself.
Use it from your agent
podcli is an MCP server, so an agent can transcribe, suggest clips, and render them through conversation.
podcli mcp install # registers it with Claude Code and Codex, whichever CLI is on PATH
--help and --version never download anything. First run of process, studio, transcribe, or multicam provisions the engine. podcli mcp does not, because stdout is JSON-RPC. Run podcli setup once before you point an agent at it.
podcli process episode.mp4 --json --yes
podcli doctor --json
--json writes one object to stdout and logs to stderr. -y/--yes skips prompts. NO_COLOR=1 (or a non-TTY) turns off colour and spinners.
rank_clip_candidates returns a shortlist so the agent does not have to read the whole transcript. process_episode runs the pipeline as one job; poll it with job_status. The workflow lives in SKILL.md. Tool names are generated into llms.txt.
Claude Desktop and Codex setup is in the MCP docs.
Content workflow
PodStack ships with podcli as a set of Claude Code slash commands. They take a transcript to a publish-ready package: scored moments, titles, descriptions, thumbnail briefs, a brand review, and a publish checklist.
/produce-shorts
The commands live in .claude/commands/. CLAUDE.md describes each one.
Docs
| Guide | What's in it |
|---|---|
| Getting started | Install, first episode, the whole flow |
| The studio | Web UI: library, episodes, content, highlights |
| CLI | Commands, flags, presets, assets |
| MCP server | Agent setup and available tools |
| Captions and formats | Styles, aspect ratios, cropping |
| Configuration | Environment variables, config profiles, transcript format |
Docs are open source at nmbrthirteen/podcli-docs.
Contributing
See CONTRIBUTING.md for the dev setup and conventions, and RELEASE.md for how releases are cut.
Credits
Content workflow powered by PodStack, inspired by gstack by Garry Tan.
License
AGPL-3.0. See LICENSE.
Need podcli without AGPL terms? A commercial license is available. Email siradze@nikusha.me with a one-line description of your use case.
podcli MCP questions
What does podcli do?
It transcribes a podcast episode, picks the moments worth clipping, crops to the visible face, and burns in captions to produce short clips.
How do I connect podcli to Claude Code or Codex?
Run podcli mcp install after a one-time podcli setup; the command registers the server with whichever CLI is on PATH.
Does podcli run locally or in the cloud?
Transcription and rendering run on your machine by default, with cloud calls only for optional services like AI scoring, AssemblyAI, or podcli Pro.
How many MCP tools does podcli expose?
It exposes 32 MCP tools that let an agent transcribe, score, and render clips through conversation.
Is podcli free and open source?
Yes, podcli is released under AGPL-3.0 with unlimited, watermark-free exports, and a commercial license is available on request.