# mediasense (pypi · mediasense)

Let Claude listen to audio and watch video: transcribe, summarize and edit recordings and links.

- Trust score: 69/100 (medium)
- Change this week: +3
- Registry status: active
- Liveness: live
- Owner verified: no
- Last scored: 2026-11-21

## Components

- pypi · `mediasense`: 69/100 (this document), [markdown](https://verifymcp.io/servers/joaoestella-mediasense/mediasense.md), [page](https://verifymcp.io/servers/joaoestella-mediasense/mediasense)

## Channel facts

- Registry: `pypi`
- Package: `mediasense`
- Version: `0.5.2`
- Transport: `stdio`

## Trust breakdown

How this component scores in each security and reliability category. Every signal is checked automatically from public evidence about the published package, including repeated runs of it in an isolated sandbox, and we only credit what we can confirm. Scores are 0–100 per category. Scoring method: https://verifymcp.io/docs/scoring (what has changed: https://verifymcp.io/docs/scoring/changelog)

Scored 2026-11-21.

- **Supply Chain Security**: 49/100
  - Malware scan not yet available for this package.
  - No known CVEs affecting this package version or its production dependencies.
  - Runs hatchling.build at install time, a recognised build step with no custom scripting around it.
  - 10 of 79 dependencies flagged as unhealthy.
- **Provenance & Transparency**: 97/100
  - Source repository is publicly reachable at the declared URL.
  - Cryptographically verified build provenance (signed, bound to joaoestella/mediasense).
  - Clear OSI-approved license (MIT).
  - Actively maintained (last published 10 days ago).
  - Disclosure check failed: no security disclosure policy was found in the source repository.
- **Schema Quality & AI Usability**: 71/100
  - 100% of prompts and resources have a non-trivial description (not blank, and not just the item's name).
  - AI-judged instruction clarity (fair).
  - Context-footprint check failed: tool/resource definitions use about 1834 tokens (~152/item across 12 items; 12 tools + 0 resources), over budget; trim descriptions and params.
  - Usage-examples check failed: none of the tools include examples.
- **Stability & Change Management**: 47/100
  - Stability observed for 14 of 30 days with no destabilising changes; credit accrues until the full window elapses.
- **Tool Coverage**: 71/100
  - 100% of tools have a non-trivial description (not blank, and not just the tool's name).
  - 0% of tool parameters carry a description.
  - Structured output schemas are declared (67% of tools); any adoption earns full credit.
- **Tool Safety**: 100/100
  - No prompt-injection markers were found in the server instructions, tool names or descriptions we captured.
  - We read all 12 captured tool definition(s), and no name or description among them implies an irreversible operation.
  - An AI judge read all 13 captured unit(s) of tool text and found none that tries to manipulate the model reading it.
- **Capabilities**: 100/100
  - Implements a current MCP spec version (2026-07-28).

## Install

### How do I install the mediasense MCP server?

mediasense runs locally as a PyPI package, launched with uvx mediasense mcp. Ready-made configuration for Claude, Cursor, VS Code, Codex and 5 more is on this page, copied from each client's own documentation.

### Claude

```bash
claude mcp add joaoestella-mediasense -- uvx mediasense mcp
```

### Cursor

```json
{
  "mcpServers": {
    "joaoestella-mediasense": {
      "command": "uvx",
      "args": [
        "mediasense",
        "mcp"
      ]
    }
  }
}
```

### VS Code

```json
{
  "servers": {
    "joaoestella-mediasense": {
      "command": "uvx",
      "args": [
        "mediasense",
        "mcp"
      ]
    }
  }
}
```

### Codex

```bash
codex mcp add joaoestella-mediasense -- uvx mediasense mcp
```

### opencode

```json
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "joaoestella-mediasense": {
      "type": "local",
      "command": [
        "uvx",
        "mediasense",
        "mcp"
      ],
      "enabled": true
    }
  }
}
```

### OpenClaw

```bash
openclaw mcp add joaoestella-mediasense --command uvx --arg mediasense --arg mcp
```

### Hermes

```yaml
mcp_servers:
  joaoestella-mediasense:
    command: "uvx"
    args: ["mediasense", "mcp"]
```

### Netclaw

```json
{
  "McpServers": {
    "joaoestella-mediasense": {
      "Transport": "stdio",
      "Command": "uvx",
      "Arguments": [
        "mediasense",
        "mcp"
      ]
    }
  }
}
```

### Vellum

```bash
assistant mcp add joaoestella-mediasense -t stdio -c uvx -a mediasense mcp
```

### Other

```json
{
  "mcpServers": {
    "joaoestella-mediasense": {
      "command": "uvx",
      "args": [
        "mediasense",
        "mcp"
      ]
    }
  }
}
```

## Changelog

Every change recorded for this component, newest first. Days that predate change tracking, or that we cannot explain, say so: "we were watching and nothing happened" and "we were not watching" are different claims.

### 2026-11-20 (score 69, +1)

No change was recorded against any check on this day. Stability & Change Management went from 40 to 43. That category is still filling its 30-day observation window: 12 days of observed history at the previous scan, 13 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-11-19 (score 68, −15)

- [security regression] Malware scan: pass → unverified

### 2026-11-18 (score 83, +1)

No change was recorded against any check on this day. Stability & Change Management went from 33 to 37. That category is still filling its 30-day observation window: 10 days of observed history at the previous scan, 11 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-11-16 (score 82, +1)

No change was recorded against any check on this day. Stability & Change Management went from 27 to 30. That category is still filling its 30-day observation window: 8 days of observed history at the previous scan, 9 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-11-15 (score 81, +15)

- [security improvement] Malware scan: unverified → pass

### 2026-11-14 (score 66, −14)

- [security regression] Malware scan: pass → unverified

### 2026-11-12 (score 80, +1)

No change was recorded against any check on this day. Stability & Change Management went from 13 to 17. That category is still filling its 30-day observation window: 4 days of observed history at the previous scan, 5 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-11-11 (score 79, +15)

- [security improvement] Malware scan: unverified → pass

## MCP tools (12)

### `check_job` (~78 tokens)

Check a running job

Get the result of a long-running mediasense job (understand_media, transcribe, render_edit, watch_video).

Waits up to ~30s; returns the finished result, or the progress so far plus the same job_id to check again.
Without job_id, lists jobs that are still running.

Input parameters:

- `job_id` (string)

### `media_info` (~31 tokens)

Media file info

Duration, resolution, fps and which streams (audio/video) a file has.

Input parameters:

- `path` (string, required)

Output parameters:

- `result` (string)

### `transcribe` (~234 tokens)

Transcribe speech

Transcribe audio/video (file path or link) with word-level timestamps; auto-detects language.

vocabulary: names/jargon likely in the audio (["Cloudflare", "Claude", "Next.js"]) so they are
spelled right. Pass them whenever you know them (from the user, the title or the screen).

model: tiny | base | small | medium | large-v3 | turbo (bigger = slower, more accurate).
backend: faster-whisper (default, local) | whisperx (local, tighter word times, speakers) | openai (API).
verbatim: keep hesitations (um, éé, hum) so they can be removed later.
diarize: label speakers (whisperx backend + HF_TOKEN env var).
Reuses an existing <file>.transcript.json unless force=true.

Input parameters:

- `backend` (string)
- `diarize` (boolean)
- `force` (boolean)
- `language`
- `model` (string)
- `path` (string, required)
- `verbatim` (boolean)
- `vocabulary`

Output parameters:

- `result` (string)

### `read_transcript` (~65 tokens)

Read a transcript

Read a transcript (pass the media path or the .transcript.json).

start/end limit the time range. words=true shows each word as word@seconds.

Input parameters:

- `end`
- `path` (string, required)
- `start`
- `words` (boolean)

Output parameters:

- `result` (string)

### `search_transcript` (~40 tokens)

Search a transcript

Find every time a word/phrase is said (case, accent and punctuation insensitive).

Input parameters:

- `path` (string, required)
- `phrase` (string, required)

Output parameters:

- `result` (string)

### `plan_edit` (~297 tokens)

Preview an edit

Preview an edit without writing anything. Same arguments as render_edit.

cut_phrases: remove these phrases (all occurrences, or only `occurrence`, 1-based).
cut_between: [["from phrase", "to phrase"], ...] removes everything from one to the other, inclusive.
cut_ranges: ["1:02-1:05.5", "90-95"].  keep_only: "0:30-2:00".
fillers: remove um/uh/éé/hum/euh/äh…; filler_words adds more ("tipo", "you know").
silences: shorten pauses longer than N seconds, leaving silence_keep seconds.
silence_source: "audio" (level based) or "transcript" (gaps between words; better with background music).
Word timestamps are approximate, so every cut point is moved to the nearest real pause in the audio
(marked "moved to the pause") and no word gets clipped. exact_cuts=true cuts at the exact times instead.

Input parameters:

- `cut_between`
- `cut_phrases`
- `cut_ranges`
- `exact_cuts` (boolean)
- `filler_words`
- `fillers` (boolean)
- `keep_only`
- `occurrence`
- `path` (string, required)
- `silence_keep` (number)
- `silence_source` (string)
- `silences`

Output parameters:

- `result` (string)

### `render_edit` (~190 tokens)

Render an edit to a new file

Apply an edit (see plan_edit for the cut arguments). Never touches the original.

Writes <file>.edited.<ext> unless media=false or output is given. edl → Premiere/Resolve/Avid,
fcpxml → Final Cut/Resolve, srt → subtitles re-timed to the edited version.

Input parameters:

- `cut_between`
- `cut_phrases`
- `cut_ranges`
- `edl` (boolean)
- `exact_cuts` (boolean)
- `fcpxml` (boolean)
- `filler_words`
- `fillers` (boolean)
- `keep_only`
- `media` (boolean)
- `occurrence`
- `output`
- `path` (string, required)
- `silence_keep` (number)
- `silence_source` (string)
- `silences`
- `srt` (boolean)

Output parameters:

- `result` (string)

### `export_transcript` (~73 tokens)

Export subtitles or text

Write the transcript as files next to the media. formats: srt, vtt, txt, csv, json.

words_per_cue (e.g. 2) makes short captions for vertical video.

Input parameters:

- `formats` (array, required)
- `path` (string, required)
- `words_per_cue`

Output parameters:

- `result` (string)

### `place_audio` (~132 tokens)

Place audio into a video

Put an audio clip (or a slice of it) into a video/audio starting at `at`.

mode: mix (both play) | duck (original lowered by duck_db) | replace (original muted meanwhile).
Video is copied without re-encoding. Writes <base>.placed.<ext> unless output is given.

Input parameters:

- `at` (string, required)
- `base` (string, required)
- `duck_db` (number)
- `insert` (string, required)
- `insert_from`
- `insert_to`
- `mode` (string)
- `output`
- `volume` (number)

Output parameters:

- `result` (string)

### `watch_video` (~90 tokens)

Key frames of a video

Understand a video: key frames at scene changes + what is said during each shot.

Returns a shot list (time, speech) and contact-sheet images labelled with shot number and time.
Transcribe first to get the speech column. Lower threshold = more sensitive scene detection.

Input parameters:

- `include_images` (boolean)
- `max_frames` (integer)
- `path` (string, required)
- `threshold`

### `get_frame` (~56 tokens)

Get one video frame

Return the video frame at `time` (e.g. "1:23.4") to look at a moment in detail.

Input parameters:

- `path` (string, required)
- `time` (string, required)
- `width` (integer)

### `understand_media` (~189 tokens)

Watch and listen to a video or audio

Watch and listen to a video or audio – a TikTok, YouTube, Instagram, X or Vimeo link, or a file –
to summarize it, explain it or answer questions about it. Use this instead of opening the link in a
browser or fetching the page (those sites block that). One call downloads the link, transcribes the
speech with timestamps and returns key-frame contact sheets with what is said in each shot, plus the
post's title/description. Long videos reply "STILL WORKING" – then call check_job.

vocabulary: brand names / jargon to spell right. language: ISO code if known (pt, en…).

Input parameters:

- `force` (boolean)
- `include_images` (boolean)
- `language`
- `max_frames` (integer)
- `model` (string)
- `source` (string, required)
- `vocabulary`

## Diagnostics

Captured diagnostic sections: Provenance, Install scripts, Dependencies. The full working is on the page: https://verifymcp.io/servers/joaoestella-mediasense/mediasense#diagnostics

## Score history

- 2026-11-21: 69
- 2026-11-20: 69
- 2026-11-19: 68
- 2026-11-18: 83
- 2026-11-17: 82
- 2026-11-16: 82
- 2026-11-15: 81
- 2026-11-14: 66
- 2026-11-13: 80
- 2026-11-12: 80
- 2026-11-11: 79
- 2026-11-10: 64
- 2026-11-09: 78
- 2026-11-08: 78
- 2026-11-07: 63

## Common questions

### What is the mediasense MCP server?

mediasense is an MCP server listed in the public MCP registry as io.github.joaoestella/mediasense. Let Claude listen to audio and watch video: transcribe, summarize and edit recordings and links. This page covers its PyPI package (mediasense).

### Is the mediasense MCP server safe to use?

mediasense scores 69 out of 100 on VerifyMCP. We found no known CVEs affecting it as of 21 November 2026. Its build provenance is signed and verified. That is a record of what we were able to check automatically, not an endorsement. The category breakdown on this page shows every signal behind the number, including the ones we could not confirm.

### What tools does the mediasense MCP server expose?

mediasense exposes 12 tools: check_job, media_info, transcribe, read_transcript, search_transcript, and 7 more. Their descriptions and schemas cost roughly 1,475 tokens of context every time the server is loaded.

### Is the mediasense MCP server still maintained?

mediasense is still listed as active in the MCP registry. We last reached this channel on 21 November 2026. Those dates come from our own scans of the registry and the channel itself, not from anything the publisher announced.

### What licence is the mediasense MCP server under?

mediasense declares the MIT licence, which is OSI-approved. That covers the source only, and says nothing about the cost of any service it calls.

### What did VerifyMCP test on the mediasense MCP server?

VerifyMCP ran 15 checks on mediasense, covering Known CVEs, Install scripts, Dependency health, Source repository and 11 more. One item could not be completed, a gap on our side rather than a finding about mediasense: Malware scan. Endpoint Security and Transport & Reachability do not apply to a PyPI package, so they were not tested. The score does not measure everything, and the scoring documentation lists what it leaves out.

## Links

- PyPI project: https://pypi.org/project/mediasense/
- Socket report: https://socket.dev/pypi/package/mediasense
- Repository: https://github.com/joaoestella/mediasense
- Website: https://joaoestella.github.io/mediasense/
- Changelog RSS feed: https://verifymcp.io/servers/joaoestella-mediasense/mediasense.xml
- Changelog JSON feed: https://verifymcp.io/servers/joaoestella-mediasense/mediasense/changelog.json
- Trust report (JSON): https://verifymcp.io/servers/joaoestella-mediasense/mediasense.json
- HTML version of this page: https://verifymcp.io/servers/joaoestella-mediasense/mediasense
