@prosodyai/mcp-docs
OfficialServer Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| PORT | No | Port for the HTTP server. | 3333 |
| PROSODYAI_REPO_ROOT | No | Path to the prosodyai monorepo root, used to locate content directory. If not set, the package walks up from its own directory. |
Capabilities
Features and capabilities supported by this server
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
| resources | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| search_docsA | Search ProsodyAI docs, SDK READMEs, recipes, and OpenAPI metadata. Returns a ranked list of matches with snippets and stable |
| list_docsA | List every document in this server. Useful for browsing without a search query. |
| read_docA | Fetch the full content of a doc, SDK README, recipe, or other entry by |
| list_endpointsA | List ProsodyAI REST API endpoints from the bundled OpenAPI spec. Optional filters by tag or path substring. |
| get_endpointA | Get the full OpenAPI operation object (parameters, request body, responses, security) for a single REST endpoint. |
| get_openapiA | Return the full bundled OpenAPI 3 spec for the ProsodyAI REST API. Use sparingly — prefer |
| list_recipesA | List curated end-to-end implementation recipes for common ProsodyAI integration tasks (e.g. add prosody to a LiveKit agent, stream from a browser, wire the LangChain tool, define KPIs). |
| get_overviewA | Return a single-page overview of the ProsodyAI platform: what it is, what to use, and how the SDKs/API/recipes relate. Read this first when starting an integration. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
| api/openapi | Structured data: openapi |
| docs/API_CUSTOM_DOMAIN | To serve the FastAPI at **https://api.prosodyai.app** and have env vars set on Cloud Run. |
| docs/KPI_EXAMPLES | Use these when defining KPIs in the dashboard or when training the KPI head. Each needs **outcome labels** (post-call / post-session) to train the model; until then the API uses heuristic mapping from prosody. |
| docs/KPI_LABELED_DATA | To train the KPI head you need **(audio or prosody features) + (actual KPI outcomes)**. Outcomes are sent via the feedback API; the same `session_id` links a conversation to its labels. |
| docs/NAMING | Single source of truth so we don’t confuse repos, packages, and the product. |
| docs/OVERVIEW | ProsodyAI is real-time prosodic intelligence infrastructure for voice agents. It tells your agent *how* someone sounds — not just what they say. Audio flows through a frozen WavLM-Large backbone into Mamba selective scan blocks that output continuous Valence-Arousal-Dominance ... |
| docs/PAPER | ProsodyAI is a speech analysis system that turns short audio chunks into affective and prosodic signals for voice agents, call analysis, and downstream business workflows. The current deployed system accepts base64-encoded audio, resamples it to 16 kHz when needed, runs a Pros... |
| docs/README | - **[STRUCTURE.md](STRUCTURE.md)** — DB, API, dashboard layout - **[SYSTEMS.md](SYSTEMS.md)** — Topology, env contract, deployment - **[PAPER.md](PAPER.md)** — Grounded technical paper for the current deployed system - **env.example** — Copy to repo root as `.env` for local de... |
| docs/schema/README | **Source of truth: `website/prisma/schema.prisma`.** Run migrations from the dashboard: |
| docs/STRUCTURE | One database. One backend API. One dashboard. **Configuration: [SYSTEMS.md](SYSTEMS.md).** |
| docs/SYSTEMS | Single source of truth for topology, configuration, and deployment. All runtime config comes from environment variables; no hardcoded URLs or secrets. |
| docs/TECHNICAL_DEEP_DIVE | 1. [System Overview](#1-system-overview) 2. [ProsodySSM Model Architecture](#2-prosodyssm-model-architecture) 3. [Training Pipeline](#3-training-pipeline) 4. [Inference & Deployment (Baseten)](#4-inference--deployment-baseten) 5. [API Layer (FastAPI on Cloud Run)](#5-api-layer... |
| docs/TRAINING | ProsodySSM training runs on **Baseten** (GPU). Data is read from **GCS**; checkpoints go to Baseten workspace (and optionally GCS). Inference uses the same Baseten stack: deploy a checkpoint as a Truss model. |
| recipes/browser-streaming | Goal: capture mic audio in the browser, stream it to the ProsodyAI realtime endpoint, and react to escalation alerts in the UI (e.g. show a "calm down" indicator, switch the agent's persona, or surface a coach card). |
| recipes/kpi-flow | ProsodyAI does **not** ship hard-coded "emotion" classes. Instead, you define the KPIs you actually care about (e.g. `retention_intent`, `clinician_handoff`, `buying_intent`, `authenticity_score`) in the dashboard, and the API returns predictions for *those* KPIs from raw pros... |
| recipes/langchain-agent | Goal: give a LangChain agent the ability to listen to an audio file or live session and reason about how the speaker sounds — separately from what they said. |
| recipes/livekit-realtime-agent | Goal: a LiveKit `Agent` that listens to the caller's audio in real time, streams it through ProsodyAI, and adapts its behaviour when the prosodic signal changes (e.g. caller becomes frustrated → switch to empathetic tone, trigger a de-escalation prompt, or hand off to a human). |
| recipes/rest-api-integration | When you can't (or don't want to) install an SDK — e.g. inside an Edge Function, a different language, or a thin proxy — call the ProsodyAI REST API directly. |
| recipes/sdk-typescript-quickstart | Use this when adding ProsodyAI to a Node, Next.js, or browser app (e.g. AureliaStudio's web client or a Vercel function). |
| sdks/api-fastapi | Public REST API for ProsodyAI speech emotion recognition service. |
| sdks/langchain | ProsodyAI integration for LangChain. Includes speech emotion analysis, forward-looking conversation predictions, and feedback for continuous model improvement. |
| sdks/livekit | The plugin can classify transcript turns alongside prosody events. Feed it text from your LiveKit STT/transcription pipeline: |
| sdks/python-core | Core prosody/emotion model library (ProsodySSM). Pip package: **prosody-ssm** (import `prosody_ssm`). |
| sdks/typescript | ProsodyAI SDK for speech emotion analysis with forward-looking conversation predictions. Supports files, buffers, real-time streaming, and feedback for continuous model improvement. |
Latest Blog Posts
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/ProsodyAI/mcp-docs'
If you have feedback or need assistance with the MCP directory API, please join our Discord server