Connect your account, then chat with AI to run tools.
Train photorealistic faces, configure PAL personas, render personalized videos, and launch real-time conversational video sessions from chat.
Live probe refreshed Sep 14, 2026 · Endpoint host mcp.mcpbundles.com
Built for
Sales Teams, Customer Success, Marketing Automation, L&D Teams, AI Product Builders
Inventory faces and personas
Grounds later video or conversation creation in live account inventory.
List my Tavus replicas and personas, summarize training status on each face, and note which persona ids have a default replica configured.
Connect Tavus to any MCP client in minutes
You’ll sign in to MCPBundles when your client connects.
https://mcp.mcpbundles.com/bundle/tavusThis server needs a sign-in. MCPBundles handles that setup once — then chat here with saved access and a log of what ran.
Opens MCPBundles Studio with this server selected. After sign-in, chat and run tools from the same thread.
Browse all toolsCheck rendered video status
Typical post-generation workflow before sharing a personalized clip.
List my recent Tavus videos, pick the newest generating or ready entry, and tell me whether a download link is available yet.
Review conversation sessions
Operators monitoring live CVI sessions without opening the Tavus dashboard.
List my Tavus conversations from the last page, show active vs ended counts, and summarize join URLs for any still-active sessions.
Plan a personalized outreach clip
Connects read-only catalog discovery to a concrete video brief.
From my Tavus replica and speech catalog, recommend which face and script pairings fit a short renewal reminder video.
What can AI agents do with Tavus on MCPBundles?
Agents can list or fetch personas, replicas, speeches, videos, and conversations; create conversations and rendered videos; update persona settings and video names; and delete personas, replicas, videos, or conversations when asked.
How do I connect Tavus to MCPBundles?
Add your Tavus API key from the Tavus developer settings on MCPBundles. The platform stores the key securely and sends it on each Tavus request.
What is the difference between a Tavus conversation and a video?
Conversations are live real-time CVI sessions with join URLs. Videos are asynchronous rendered clips from a replica plus script or audio — use the video wait tool or poll get until status is ready before sharing download links.
Domain knowledge for Tavus — workflow patterns, data models, and gotchas for your AI agent.
Tavus builds personalized video and real-time conversational video (CVI). Modern API terms are PAL (behavior, prompts, pipeline) and Face (photorealistic likeness). Legacy names persona and replica still work on the same endpoints.
Tavus Delete Conversation
Permanently delete a Tavus conversation and associated data by id. Destructive and idempotent. Pass hard=true only when irrevocable removal is intended.
Tavus Delete Persona
Permanently delete a Tavus persona (PAL) by id.
Tavus Delete Replica
Delete a Tavus replica (face) by id. Deleted replicas cannot be used in new conversations or videos.
Tavus Delete Video
Delete a rendered Tavus video by id. Pass hard=true only when irrevocable asset removal is intended.
Tavus Get Conversations
List or fetch Tavus real-time video conversations (CVI sessions). Omit id to page through conversations; pass id for one session including join URL, status, face and PAL ids, and optional verbose event payloads when verbose=true.
Tavus Get Personas
List or fetch Tavus personas (PALs) that define behavior, prompts, and default faces for conversational video. Omit id to list; pass id for one persona including system prompt, default face, and layer configuration.
Tavus Get Replicas
List or fetch Tavus replicas (photorealistic faces) used in videos and live conversations. Omit id to list training status and thumbnails; pass id for one replica including model name and training progress. Use the replica wait tool to block until training completes after create.
Tavus Get Speeches
List speech scripts available in your Tavus account for personalized video generation. The upstream speech catalog is list-only; pick a script id from the response when creating videos.
Tavus Get Videos
List or fetch Tavus rendered videos. Omit id to page through generation status, scripts, and download links; pass id for one video including ready-state URLs when generation completed. Use the video wait tool to block until ready after create.
Tavus Upsert Conversation
Start a new real-time Tavus video conversation. Creation only — pass face_id and pal_id (or a PAL with a default face). Returns conversation_id, join URL, and status. Use test_mode=true to validate payload without starting a billable session.
Tavus Upsert Persona
Create or update a Tavus persona (PAL). Omit id to POST a new persona — requires pal_name and system_prompt unless using echo pipeline mode. Pass id to PATCH common fields (pal_name, system_prompt, default_replica_id) on the live persona.
Tavus Upsert Replica
Create a Tavus replica (face) from training media. Creation only — training is asynchronous. Provide train_video_url or train_image_url (with voice_name when using image training). Returns replica id and training status.
Tavus Upsert Video
Create a rendered Tavus video or rename an existing one. Omit id to POST a new video — requires replica_id and either script or audio_url. Pass id with video_name to rename an existing video (PATCH name only).
Tavus Wait For Replica
Poll a Tavus replica (face) until training completes, fails, or times out. Use after creating a replica when status is started.
Tavus Wait For Video
Poll a Tavus rendered video until status is ready, error, deleted, or times out. Use after creating a video when status is queued or generating.
Agents can list or fetch personas, replicas, speeches, videos, and conversations; create conversations and rendered videos; update persona settings and video names; and delete personas, replicas, videos, or conversations when asked.
Add your Tavus API key from the Tavus developer settings on MCPBundles. The platform stores the key securely and sends it on each Tavus request.
Conversations are live real-time CVI sessions with join URLs. Videos are asynchronous rendered clips from a replica plus script or audio — use the video wait tool or poll get until status is ready before sharing download links.
Replica training is asynchronous. New faces stay in started status until training completes; list or fetch the replica row again before starting conversations or videos.
Add the MCPBundles server URL to your MCP client configuration (Claude Desktop, Cursor, VS Code, etc.). The URL format is: https://mcp.mcpbundles.com/bundle/tavus. Authentication is handled automatically.
Tavus provides 15 tools that can be called by AI agents, along with a SKILL.md that gives your AI agent domain knowledge about when and how to use them.
Tavus uses API Key. Tavus requires credentials. Connect via MCPBundles and authentication is handled automatically.
© 2026 ThinkChain Inc. All rights reserved.
MCPBundles is an independent platform built on the open Model Context Protocol standard. Not affiliated with Anthropic PBC or Claude.
Other MCP servers in this category from the directory index
Vercel exposes 5 public tools on this MCP endpoint.
5 toolsAI design tool to turn prompts into production-ready UI. Use your design system, generate prototypes...
Splice is the industry's highest quality, royalty-free sample library. Get access to millions of sou...
Canva MCP Server enables AI assistants to create designs, autofill templates, manage folders, search...