Reference
MCP tools
Reference for all eight YouTube Translate MCP tools, including parameters, return shapes, and the budget rule for paid work.
The server exposes eight tools. Five of them return content; three manage processing.
The budget rule
Uncached work costs money, so it takes two steps.
get_processing_quotereturnsmaximumCreditUnitswithout spending anything.- The caller shows that number to the user and gets approval.
- The content tool is called with
max_credit_unitsset to at least the quoted maximum.
If a video is already cached, none of this applies: the read returns immediately at the cached price and max_credit_units is ignored.
Calling a content tool on an uncached video without max_credit_units is refused rather than silently charged.
Content tools
get_transcript
Returns the spoken words, with optional timestamps.
| Parameter | Type | Required | Notes |
|---|---|---|---|
url |
string | yes | Public YouTube URL |
include_timestamps |
boolean | no | Adds segment timestamps |
max_credit_units |
integer | for uncached | Must be at least the quoted maximum |
get_translation
Renders a retained transcript into another language.
| Parameter | Type | Required | Notes |
|---|---|---|---|
url |
string | yes | Public YouTube URL |
language |
string | yes | Two-letter code, for example es |
max_credit_units |
integer | for uncached | Quoted maximum or higher |
get_subtitles
Returns an SRT or VTT file. Generates the translation first if that language is missing.
| Parameter | Type | Required | Notes |
|---|---|---|---|
url |
string | yes | Public YouTube URL |
language |
string | no | Defaults to en |
format |
string | no | srt or vtt, defaults to srt |
max_credit_units |
integer | for uncached | Quoted maximum or higher |
get_summary
Condenses the transcript at one of three lengths.
| Parameter | Type | Required | Notes |
|---|---|---|---|
url |
string | yes | Public YouTube URL |
language |
string | no | Defaults to en |
length |
string | no | short, medium, or long |
max_credit_units |
integer | for uncached | Quoted maximum or higher |
search_video
Finds a topic or quotation inside one video and returns matching passages with timestamps.
| Parameter | Type | Required | Notes |
|---|---|---|---|
url |
string | yes | Public YouTube URL |
query |
string | yes | What to look for |
language |
string | no | Defaults to en |
max_credit_units |
integer | for uncached | Quoted maximum or higher |
Processing tools
get_processing_quote
Returns the cost ceiling and your current balance. Starts nothing.
| Parameter | Type | Required | Notes |
|---|---|---|---|
url |
string | yes | Public YouTube URL |
operation |
string | no | transcribe, translate, or summarize; defaults to transcribe |
language |
string | no | Defaults to en |
length |
string | no | Summary length, defaults to medium |
start_video
Queues processing and returns a job ID without waiting. Use this for long videos so the conversation is not blocked.
| Parameter | Type | Required | Notes |
|---|---|---|---|
url |
string | yes | Public YouTube URL |
max_credit_units |
integer | for uncached | Quoted maximum or higher |
get_video_status
Reports whether processing finished, failed, or is still running.
| Parameter | Type | Required | Notes |
|---|---|---|---|
video_id |
string | yes | The ID returned by start_video |
Typical sequences
Cached video, lowest cost
get_transcript(url)
Uncached video, with approval
get_processing_quote(url, operation="transcribe")
→ user approves maximumCreditUnits
get_transcript(url, max_credit_units=<approved>)
Long video, non-blocking
get_processing_quote(url)
start_video(url, max_credit_units=<approved>)
get_video_status(video_id) # repeat until completed
get_transcript(url)
The last call is a cached read, because start_video already produced the artifact.
Related
- Pricing and credits for what each tool costs
- Errors and retries for failure handling
- Limits and retention for source caps