WAVE MCP Server

Official

Overview Schema Source Related Servers Discussions Admin

wave_start_captions

Transcribe a single audio clip and optionally run a fast LLM step (summarize, moderate, translate, custom) over the resulting transcript. Send base64 audio to WAVE's live pipeline to generate captions or translations.

Instructions

Transcribe an audio clip and optionally run a fast-LLM step over the transcript (POST /v1/live/pipeline, multipart). This processes ONE provided audio chunk through WAVE's live pipeline — it does not attach a persistent caption feed to a live stream. Requires the account's live pipeline feature to be enabled; a 404 means it is not armed for this account

Input Schema

TableJSON Schema

Name	Required	Description
`mode`	No	Fast-LLM step to run over the transcript (default: summarize)
`task`	No	Caption task (default: transcribe)
`model`	No	Transcription model (default: whisper-large-v3-turbo; translate forces whisper-large-v3)
`filename`	No	Filename hint for the audio (default: "audio.wav")
`language`	No	ISO 639 language hint for transcription (transcribe task only)
`llm_model`	Yes	Fast LLM model ID to run the pipeline step
`stream_id`	No	Client correlation ID for this stream/session (letters, digits, . _ : -, 1-128 chars)
`max_tokens`	No	Max tokens for the LLM step (default: 256)
`instruction`	No	Custom instruction for mode=custom
`audio_base64`	Yes	Base64-encoded audio bytes to transcribe (max 25MB decoded)

Schema Changelog

Changes observed during successful MCP inspections.

Changed17 schema fields changedv0.3.0Sep 6, 2026

removedInput schema / additionalProperties
```
Removed value: -false
```

addedInput schema / properties / audio_base64

Added value: +{
+  "description": "Base64-encoded audio bytes to transcribe (max 25MB decoded)",
+  "minLength": 1,
+  "type": "string"
+}

addedInput schema / properties / filename

Added value: +{
+  "description": "Filename hint for the audio (default: \"audio.wav\")",
+  "type": "string"
+}

addedInput schema / properties / instruction

Added value: +{
+  "description": "Custom instruction for mode=custom",
+  "maxLength": 500,
+  "type": "string"
+}

changedInput schema / properties / language / description

Previous value: -"ISO 639-1 language code (default: en)"New value: +"ISO 639 language hint for transcription (transcribe task only)"

removedInput schema / properties / language / maxLength
```
Removed value: -2
```
removedInput schema / properties / language / minLength
```
Removed value: -2
```

addedInput schema / properties / llm_model

Added value: +{
+  "description": "Fast LLM model ID to run the pipeline step",
+  "minLength": 1,
+  "type": "string"
+}

addedInput schema / properties / max_tokens

Added value: +{
+  "description": "Max tokens for the LLM step (default: 256)",
+  "maximum": 4096,
+  "minimum": 1,
+  "type": "integer"
+}

addedInput schema / properties / mode

Added value: +{
+  "description": "Fast-LLM step to run over the transcript (default: summarize)",
+  "enum": [
+    "summarize",
+    "moderate",
+    "translate",
+    "custom"
+  ],
+  "type": "string"
+}

addedInput schema / properties / model

Added value: +{
+  "description": "Transcription model (default: whisper-large-v3-turbo; translate forces whisper-large-v3)",
+  "enum": [
+    "whisper-large-v3-turbo",
+    "whisper-large-v3"
+  ],
+  "type": "string"
+}

removedInput schema / properties / provider

Removed value: -{
-  "description": "Transcription provider (default: deepgram)",
-  "enum": [
-    "deepgram",
-    "assemblyai",
-    "cohere"
-  ],
-  "type": "string"
-}

changedInput schema / properties / stream_id / description

Previous value: -"The stream ID"New value: +"Client correlation ID for this stream/session (letters, digits, . _ : -, 1-128 chars)"

removedInput schema / properties / stream_id / format
```
Removed value: -"uuid"
```
addedInput schema / properties / stream_id / maxLength
```
Added value: +128
```

addedInput schema / properties / task

Added value: +{
+  "description": "Caption task (default: transcribe)",
+  "enum": [
+    "transcribe",
+    "translate"
+  ],
+  "type": "string"
+}

changedInput schema / required

Previous value: -[
-  "stream_id"
-]New value: +[
+  "audio_base64",
+  "llm_model"
+]

First observedv0.1.5Apr 19, 2026

TDQS

A4.2/5.0

Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations, the description carries the behavioral burden, and it does so well: it discloses the one-shot nature, the non-persistent behavior, the endpoint/multipart detail, and the 404 meaning when the feature is not armed. It does not mention response format or potential side effects like usage costs, but the core behavioral traits are disclosed.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Two sentences, no filler. The main purpose is front-loaded, the scope constraint and non-persistent clarification come immediately after, and the operational prerequisite/error semantic is compactly appended.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness4/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a 10-parameter tool with no annotations and no output schema, the description covers the key operational constraints: single chunk, non-persistent feed, feature prerequisite, and 404 diagnosis. It is complete enough to guide selection and invocation, though it omits explicit return-value expectations and sibling comparisons.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 100%, so the schema already documents all 10 parameters. The description adds context by framing the call as processing one audio chunk and running a fast-LLM step, which helps with audio_base64 and llm_model, but it does not go beyond that baseline.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description states a specific verb ('Transcribe') and resource ('an audio clip') and immediately clarifies the optional fast-LLM step over the transcript. It also distinguishes itself from a persistent caption feed, which is important given the 'captions' name and sibling streaming tools.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description gives clear context: it processes exactly ONE provided audio chunk and explicitly says it does not attach a persistent caption feed, so an agent knows not to use it for ongoing live-stream captioning. It also provides a prerequisite (live pipeline feature enabled) and explains the 404 failure mode, but it does not name specific alternative sibling tools.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

wave_start_captions

Instructions

Input Schema

Schema Changelog

TDQS

Other Tools