Developer Tools

Tavus

by Tavus

Overview

Build and manage Tavus PALs through conversation. Choose available faces and voices, configure behavior and knowledge, run text tests, and open a video preview. Connect your Tavus account to access its resources and existing entitlements. Includes the same tools as the hosted Tavus MCP server. Custom face training, rendered video generation, PAL sales, and ownership transfers are not exposed by this integration.

Tools

tavus_chat_end

ChatGPT

End a CVI text-mode conversation.

tavus_chat_start

ChatGPT

Start a CVI text-mode conversation for testing a PAL without rendering video.

tavus_chat_turn

ChatGPT

Send one text turn to a chat-mode conversation (from tavus_chat_start) and wait for the PAL's reply over RQH's HTTP respond bridge. The first turn after starting a conversation can take extra time while the replica boots.

tavus_conversation_create

ChatGPT

Create a Tavus conversation for a pal and/or face.

tavus_conversation_end

ChatGPT

End a Tavus conversation.

tavus_deployment_create

ChatGPT

Create a live deployment that puts a PAL in front of people. channel is 'landing-page' (a page Tavus hosts, with its own link), 'embed' (the conversation inside a page of the user's site) or 'widget' (a floating launcher on every page of the user's site). The deployment is active as soon as it exists, so check tavus_deployment_list for one to reuse, then confirm the channel and every value below with the user before calling. unique_memory_per_visitor: ask the user whether each visitor gets their own memory of past calls. Embed and widget: ask where the site is hosted and pass those origins in allowed_origins, for example ["https://www.example.com"]. An empty list lets any site load it. Landing page: draft page_title and page_description from what the PAL does and show them to the user. page_cta adds outbound buttons; use only URLs the user gave. Limits: ask the user how much use to allow. Without them a public deployment has no cap on calls, and every call is billed. daily_call_limit caps calls per day across all visitors, daily_call_limit_per_user per visitor, concurrent_call_limit calls at the same moment, and max_call_duration the length of one call in seconds (10 minutes is 600). Leave one out for no limit. context is what every call from this deployment should know, such as a location or an event. name defaults to "<PAL name> Link", "Embed" or "Widget". The response carries an integration block with the landing page link, or the snippet, the coding-agent prompt to follow and the PAL's page tools.

tavus_deployment_get

ChatGPT

Get one deployment with its integration block: the landing page link, or the snippet, the coding-agent prompt to follow and the PAL's current page tools for an embed or widget.

tavus_deployment_list

ChatGPT

List deployments, newest first. Pass pal_id to see how a PAL is already deployed before creating another deployment for it.

tavus_describe_pal_options

ChatGPT

Describe one pal plus account resources that can be attached to it.

tavus_document_chunks

ChatGPT

Read extracted Knowledge chunks indexed for a document.

tavus_document_create

ChatGPT

Create a Knowledge document from an already-hosted URL. Hosted MCP cannot read local files; use the local tavus-mcp server for file-path uploads.

tavus_document_get

ChatGPT

Get one Knowledge document record.

tavus_document_list

ChatGPT

List Knowledge documents in the selected account.

tavus_document_recrawl

ChatGPT

Trigger recrawl for a crawl-backed Knowledge document.

tavus_document_tags

ChatGPT

List document tags available for Knowledge attachment.

tavus_face_list

ChatGPT

List faces, optionally restricted to stock/system faces.

tavus_guardrail_create

ChatGPT

Create a first-class guardrail.

tavus_guardrail_delete

ChatGPT

Delete a guardrail.

tavus_guardrail_get

ChatGPT

Get one guardrail.

tavus_guardrail_list

ChatGPT

List account or system guardrails.

tavus_guardrail_patch

ChatGPT

Update a first-class guardrail. Omitted fields remain unchanged.

tavus_guardrail_tags

ChatGPT

List tags used by account guardrails.

tavus_mcp_health_check

ChatGPT

Check MCP server config, auth, and optionally a downstream Tavus API read probe.

tavus_objective_create

ChatGPT

Create a validated objective set.

tavus_objective_delete

ChatGPT

Delete an objective set.

tavus_objective_get

ChatGPT

Get one objective set.

tavus_objective_list

ChatGPT

List objective sets that can be attached to PALs.

tavus_objective_patch

ChatGPT

Patch an objective set with JSON Patch operations.

tavus_objective_validate

ChatGPT

Validate an objective set without persisting it.

tavus_pal_build_and_verify

ChatGPT

Build and publish a PAL through the guided PAL Maker pipeline from a completed brief and optional interview Q&A. Drafts the personality, objectives, guardrails, and greeting, selects a face, publishes the PAL, runs conversation probes, and returns the PAL ID, editor URL, and judged verification result. The briefs fields describe each artifact, including knowledge, tools, skills, perception, and face preferences. Connected tools or knowledge sources may access external services during building or testing. Usually takes one to three minutes.

tavus_pal_capabilities_catalog

ChatGPT

List PAL Maker capabilities and their RQH mappings.

tavus_pal_capabilities_list

ChatGPT

List Magic Canvas, Slide Presenter, Web Search, Perception, and Memory attachment state for a PAL.

tavus_pal_capability_attach

ChatGPT

Enable a built-in capability by intent. Use magic_canvas for scorecards, quizzes, forms, charts, and on-screen text; pass magic_canvas_components and magic_canvas_usage_guidance in this call. Use slide_presenter for decks, web_search for current information, perception for screen/camera awareness, and memory for returning participants. Read tavus_pal_capabilities_list afterward to verify the saved config. When the creator describes a change to an existing PAL rather than an exact setting, use tavus_pal_build_and_verify with pal_id and briefs.skills instead.

tavus_pal_capability_detach

ChatGPT

Detach a PAL Maker capability. Perception is detached by setting perception_model to off.

tavus_pal_capability_patch

ChatGPT

Update an enabled built-in capability. For Magic Canvas, keep display choreography in magic_canvas_usage_guidance. The patch merges top-level config fields but replaces config.components when supplied. Read the saved config and preserve unchanged component settings, including any booking link, when changing components. Read tavus_pal_capabilities_list afterward to verify the saved config.

tavus_pal_create

ChatGPT

Create a PAL directly from supplied configuration and return its PAL Maker editor URL. This low-level operation does not draft or verify the PAL through the guided PAL Maker pipeline; for a new PAL, use tavus_pal_build_and_verify unless the creator hands over a finished system prompt and asks to skip the build. Full-pipeline PALs default to tavus-gemma-4 unless layers.llm.model is supplied. Tools are first-class records referenced by top-level tool_ids; inline layers.llm.tools configuration is unsupported.

tavus_pal_delete

ChatGPT

Delete a pal by ID.

tavus_pal_from_template

ChatGPT

Create a pal from a built-in prompt template.

tavus_pal_get

ChatGPT

Get one pal by ID.

tavus_pal_knowledge_add

ChatGPT

Attach existing Knowledge documents or document tags to a PAL.

tavus_pal_knowledge_list

ChatGPT

Show Knowledge documents, tags, resolved document records, and RAG settings attached to a PAL.

tavus_pal_knowledge_remove

ChatGPT

Remove Knowledge documents or document tags from a PAL.

tavus_pal_knowledge_replace

ChatGPT

Replace the PAL's Knowledge document and tag selections.

tavus_pal_knowledge_settings

ChatGPT

Patch RAG tuning options under layers.knowledge_base.

tavus_pal_list

ChatGPT

List compact PAL summaries, oldest first by creation time. Set sort to "desc" for the newest PALs first. Use tavus_pal_get for full settings.

tavus_pal_preview

ChatGPT

Start a full audio/video preview conversation for a PAL.

tavus_pal_skill_attach

ChatGPT

Attach or replace one raw RQH skill on a PAL.

tavus_pal_skill_detach

ChatGPT

Detach one raw RQH skill from a PAL.

tavus_pal_skill_patch

ChatGPT

Patch one raw RQH skill config attached to a PAL.

tavus_pal_skills_list

ChatGPT

List raw RQH skills attached to a PAL.

tavus_pal_tools_attach

ChatGPT

Attach first-class tools to a PAL by tool_id. Vision and audio tools automatically enable Raven perception when needed.

tavus_pal_tools_detach

ChatGPT

Detach one first-class tool from a PAL.

tavus_pal_tools_list

ChatGPT

List first-class tools attached to a PAL.

tavus_patch_pal

ChatGPT

Patch a PAL with validated JSON Patch operations for an exact field edit. To change what a PAL does or shows, such as adding an on-screen scorecard, use tavus_pal_build_and_verify with pal_id instead. Set wake_phrase and sleep_phrase under /layers/conversational_flow. Patch /layers/conferencing separately from behavior settings; it accepts username, allowlist, and whitelabel. Configuring an identity does not join a call. Ask the creator for the meeting link before joining. Attach first-class tools through top-level tool_ids or tavus_pal_tools_attach; legacy layers.llm.tools writes are rejected.

tavus_pronunciation_dictionary_create

ChatGPT

Create a pronunciation dictionary.

tavus_pronunciation_dictionary_delete

ChatGPT

Delete a pronunciation dictionary.

tavus_pronunciation_dictionary_get

ChatGPT

Get one pronunciation dictionary.

tavus_pronunciation_dictionary_list

ChatGPT

List pronunciation dictionaries that can be assigned through layers.tts.pronunciation_dictionary_id.

tavus_pronunciation_dictionary_patch

ChatGPT

Update a pronunciation dictionary. Supplying rules replaces the full rule list.

tavus_quickstart

ChatGPT

Create a PAL from a complete system prompt, select a stock face when none is supplied, and start a conversation. Returns the PAL, conversation, and selected face. This quick setup does not gather an interview or run the guided PAL Maker build and verification pipeline; use it only when the creator hands over a finished system prompt and asks to skip the build, and otherwise use tavus_pal_build_and_verify.

tavus_resource_list

ChatGPT

List account resources: guardrails, objectives, documents, voices, skills, or tools.

tavus_scaffold_embed

ChatGPT

Return starter files for embedding a Tavus conversation URL.

tavus_skill_get

ChatGPT

Get one RQH skill definition by ID.

tavus_skill_list

ChatGPT

List RQH skills that can be attached to PALs.

tavus_tool_create

ChatGPT

Create a saved tool for a real external action or application handler. Check built-in capabilities first: scorecards, quizzes, forms, charts, and on-screen text use tavus_pal_capability_attach with magic_canvas. A saved tool cannot render Magic Canvas. Backend recording is separate from on-screen display. Use web_search for weather or current information. Never invent an endpoint or placeholder handler. Attach the returned tool_id to a PAL with tavus_pal_tools_attach or top-level tool_ids. When the tool is part of building or changing a PAL, tavus_pal_build_and_verify (with pal_id for an existing PAL) drafts saved tools and built-ins together through briefs.tools and briefs.skills.

tavus_tool_delete

ChatGPT

Delete a first-class tool.

tavus_tool_get

ChatGPT

Get one first-class tool.

tavus_tool_list

ChatGPT

List compact first-class tool summaries. Omit limit to include every API page, or set limit and page to fetch one page. Use tavus_tool_get for the full definition.