10 Best AI Video MCP Servers for Claude, ChatGPT and Cursor in 2026
2026-10-20 · blog comparison / pricing · Higgsfield · original
Topic: Prompting & AI assistants (also: Higgsfield platform & ecosystem, Cost, credits & plans, Audio, voice & music, Post-production & finishing) · mode: Review & comparison · tags: flattened table
Higgsfield's blog of 20 October 2026 compares ten MCP servers that let Claude, ChatGPT, Cursor or coding agents generate video, by supported clients, sign-in, billing and handling of long renders. Higgsfield MCP uses one endpoint across clients, needs a paid plan and spends plan credits on every generation, separate from the Higgsfield API's own billing. The practical advice: have the agent state the credit cost before generating, follow long renders with a status tool rather than a second generate call, and pick the server by client and workflow.
What is said
- Frequently asked questions
What is said
★ — how much it matters to the source.
- ★★★ Higgsfield's blog compares ten MCP servers for generating video from Claude, ChatGPT, Cursor or coding agents; they differ in documented clients, whether an API key is needed, and whose balance pays. · Higgsfield MCP, Claude, ChatGPT, Cursor
- ★ Higgsfield's MCP server comparison rests on each vendor's own MCP pages and developer docs, with facts verified on 17 October 2026. · Higgsfield MCP
- ★★ Higgsfield's blog checked every video MCP server on the same criteria: documented clients, models, setup and sign-in, billing, handling of long renders, and features beyond generation.
- ★★ Higgsfield ranks Higgsfield MCP first because one connector covers a multi-model catalog with characters, audio and ad workflows on one account and one plan-credit balance, with no API key. · Higgsfield MCP, credit
- ★★★ Higgsfield MCP's one endpoint, https://mcp.higgsfield.ai/mcp, works as a custom connector in Claude, an official plugin in ChatGPT and a marketplace add-on in Cursor. · Higgsfield MCP, Claude, ChatGPT, Cursor, Higgsfield ChatGPT Plugin
- ★★ Higgsfield MCP offers 35+ models and tools, including Seedance 2.5, Kling 3.0, Veo 3.1, Wan 3.0, Cinema Studio 4.0 and MiniMax H3. · Higgsfield MCP, Seedance 2.5, Kling 3.0, Veo 3.1, Wan 3.0, Cinema Studio 4.0
- ★★★ Higgsfield MCP bills the plan credits of the signed-in account and is not connected to Higgsfield API billing, which runs on its own key and a prepaid dollar balance. · Higgsfield MCP, Higgsfield API, credit
- ★★ For long renders Higgsfield MCP returns job IDs and has a long-poll tool that tracks up to twelve jobs. · Higgsfield MCP
- ★★ Beyond generation, Higgsfield MCP offers Soul characters, reference Elements, voiceover and dubbing, Marketing Studio, Ad Multiplier, Genjutsu and skills. · Higgsfield MCP, Soul ID, Elements, Marketing Studio, Genjutsu
- ★★★ Higgsfield MCP requires a paid subscription, and every generation through the connector uses credits. · Higgsfield MCP, credit
- ★★★ Higgsfield Unlimited and free generations apply only on higgsfield.ai, so every Higgsfield MCP generation uses credits. · Higgsfield Unlimited, Higgsfield MCP, credit
- ★★ Through Higgsfield in ChatGPT, audio generation and the Website Building skill are not available. · Higgsfield MCP, ChatGPT, Higgsfield ChatGPT Plugin
- ★★ Higgsfield MCP has no hard credit cap per session; spending limits are an instruction given to the agent in the prompt. · Higgsfield MCP, credit
- ★★★ For Higgsfield MCP, Higgsfield advises asking the agent to state the credit cost and wait for your confirmation before generating. · Higgsfield MCP, credit
- ★★ Runway MCP connects Claude.ai, Claude Desktop, ChatGPT, Grok Bot, Cursor and Replit to a Runway account; it installs as a plugin in ChatGPT, Grok Bot and Cursor, elsewhere by URL, with no API key. · Claude, ChatGPT, Cursor
- ★★ Runway MCP offers Gen-4.5 plus Seedance 2.5, Kling 3.0, Veo 3.1, GPT Image 2 and Nano Banana Pro, depending on the Runway plan. · Gen-4.5, Seedance 2.5, Kling 3.0, Veo 3.1, GPT Image 2, Nano Banana Pro
- ★★ Runway's Explore Mode is not available through Runway MCP, so every generation there uses Runway credits.
- ★★ Polling is undocumented for Runway MCP, which matters for renders that outlast a client timeout.
- ★★ Higgsfield's blog calls Runway MCP best for Runway subscribers who want their existing plan inside agents.
- ★★ HeyGen Remote MCP is a hosted endpoint documented for Claude, Codex CLI, Cursor, Gemini CLI, Lovable, Manus, Superhuman and OpenAI, with an Add to ChatGPT button; it uses OAuth and every HeyGen plan qualifies. · Claude, Claude Code, Codex, Cursor, ChatGPT
- ★★ HeyGen Remote MCP offers HeyGen avatars plus one native scene model, heygen-video-1. · AI avatar
- ★★ Through HeyGen Remote MCP, HeyGen's Video Agent scripts, casts and renders from one prompt, and the scene model heygen-video-1 takes text, an image or up to twelve references.
- ★★ Beyond generation, HeyGen Remote MCP offers Video Agent, voice cloning, translation, lipsync, brand kits and batches of up to 100 jobs. · voice cloning, lip sync
- ★ Custom agents on a domain HeyGen has not approved need whitelisting to use HeyGen Remote MCP.
- ★★ HeyGen Remote MCP's toolset is built around presenters, with one scene model for clips without an avatar. · AI avatar
- ★★ Higgsfield's blog calls HeyGen Remote MCP best for teams producing presenter-led or translated videos at batch scale.
- ★★ fal.ai MCP's guides for Claude, Claude Code, Codex CLI and Cursor use an OAuth relay with browser sign-in and no key; ChatGPT gets the same server as a plugin. · Claude, Claude Code, Codex, Cursor, ChatGPT
- ★ fal.ai's MCP overview page still shows a standard endpoint with a FAL_KEY in the client config for Claude Code, Claude Desktop, Cursor and Windsurf. · Claude Code, Cursor
- ★★ fal.ai MCP is a free server, and each model run is billed at fal's direct API prices.
- ★★ fal.ai MCP allows 2 concurrent requests on a new account, rising to 40 with credit purchases.
- ★★★ For video on fal.ai MCP, fal's docs advise using submit job then check job from the start and not resubmitting to check progress, since each resubmission is a new billable job.
- ★★ fal.ai MCP returns raw outputs, so characters, audio tracks and ad formats are assembled by the user.
- ★★ Higgsfield's blog calls fal.ai MCP best for developers who want the fal catalog inside an agent and control over the queue.
- ★★ Dora's hosted MCP server serves ChatGPT, Claude, Cursor and Google Antigravity, signs in by OAuth 2.1 or a static API key, and exposes four tools: list models, generate video, generate image and check job. · ChatGPT, Claude, Cursor
- ★★ Dora offers 16 models over MCP, 9 of them video, including Grok Imagine 1.5, Seedance 1.5 Pro and 2, Kling 2.6 and Veo 3 Fast. · Seedance 2.0, Kling 2.6, Veo 3
- ★★ Dora's MCP generations are paid in coins on a paid plan or credit pack; there is no free tier.
- ★★ Through MCP, Dora offers motion transfer with Kling 2.6 and native audio on four models. · Kling 2.6, motion transfer, native audio
- ★★ Dora's list models returns the coin cost of every model, cheapest first, and coins are refunded automatically when a job fails.
- ★★ Dora's listed costs are defaults, so longer or higher-resolution clips cost more than list models shows.
- ★★ Dora has no public REST API, so access is through its web app, iPhone app or MCP.
- ★★ Higgsfield's blog calls Dora best for agents that need to compare and approve cost before each generation.
- ★★ Vivideo MCP runs hosted, with an API key in the Authorization header, or locally as the open-source MIT npm package @vivideo/mcp with the same 13 tools. · Claude Code, Cursor
- ★★ Vivideo MCP offers 32 manual-mode models plus the vivideo-auto pipeline.
- ★★ Vivideo MCP bills by usage, per second, per model and per resolution, and refunds failed renders.
- ★★ Vivideo MCP's catalog lists Seedance 2.0, with newer Seedance versions not yet in its API table. · Seedance 2.0
- ★★ Vivideo MCP has a wait for video tool and estimate credits before a run; on the hosted server each wait call is capped at about 20 seconds and returns timed out so the agent calls again.
- ★★ Vivideo MCP's auto mode produces a 5 to 300 second avatar video with script, voiceover and captions. · AI avatar
- ★★ Vivideo documents nine agent skills that teach an assistant to check credits first and avoid duplicate renders.
- ★★ Vivideo MCP keeps its API key in the config file and has no OAuth sign-in.
- ★★ Higgsfield's blog calls Vivideo MCP best for builders who want an auditable open-source server with avatars and voices.
- ★★ AITuber's hosted MCP server is documented for Claude, ChatGPT through a developer-mode custom connector, Cursor, Codex, Notion AI, OpenClaw and Hermes; sign-in by account or an API key header for unattended agents. · Claude, Claude Code, ChatGPT, Cursor, Codex, Notion, OpenClaw, Hermes Agent
- ★★ In AITuber's full videos the pipeline handles script, voice, visuals, captions and assembly, so a different model cannot be picked per shot.
- ★★ AITuber's MCP server is free, videos are charged in AITuber credits, and signup gives 100 free credits. · credit
- ★★ Beyond generation, AITuber offers 1,300+ voices, music videos, character and avatar templates, reusable elements, export and publishing to YouTube, TikTok, Instagram, Facebook, Threads and X. · YouTube, TikTok, Instagram
- ★★ AITuber's MCP server has a create clip tool for one standalone clip of 1 to 30 seconds without narration, and a clip-models endpoint listing models with their costs.
- ★★ AITuber videos run from 15-second Shorts to 20 minutes.
- ★★ Higgsfield's blog calls AITuber best for channels that want a narrated, captioned video published from one prompt.
- ★★ OpenArt MCP is documented for Claude as a custom connector, ChatGPT as a plugin activated with @OpenArt, Cursor through mcp.json, Kimi as a plugin and other agents over Streamable HTTP; OAuth, no API key. · Claude, ChatGPT, Cursor
- ★★ OpenArt MCP's model list includes Kling 3.0 Omni, Grok Imagine 1.5, Seedance 2.0, Wan 2.7 and PixVerse V6. · Seedance 2.0, Wan 2.7
- ★★ OpenArt's Character and Smart Shot are on its roadmap and listed as unavailable through OpenArt MCP for now.
- ★★ Managed workspaces on Claude Team or ChatGPT Business need an admin to add the OpenArt MCP connector first. · Claude, ChatGPT
- ★★ When several MCP servers are connected, naming OpenArt in the prompt is recommended so the agent uses OpenArt MCP.
- ★★ Higgsfield's blog calls OpenArt MCP best for teams that already keep their assets and projects in OpenArt.
- ★★ Imagine.art MCP sets up in eleven clients, including Claude, Claude Code, Cursor, Codex, Hermes, OpenClaw and Grok, with two commands; ChatGPT uses a directory plugin, Claude web only the URL; OAuth. · Claude, Claude Code, Cursor, Codex, Hermes Agent, OpenClaw, ChatGPT
- ★★ Imagine.art MCP's catalog covers image, video and music, with images up to 4K and music tracks of around thirty seconds.
- ★★ Imagine.art MCP clips run five or ten seconds, depending on the model.
- ★★ Of Imagine.art MCP's 96 tools, seven cover common requests and the other 89 are job-specific, from UGC avatars to 3D logo animation.
- ★★ Beyond generation, Imagine.art MCP offers Ad Studio, Fashion Studio, product and brand pipelines, upscaling and background removal.
- ★★ Some Imagine.art MCP pipelines need desktop apps such as Photoshop or After Effects installed locally. · Photoshop, After Effects
- ★★ Higgsfield's blog calls Imagine.art MCP best for commerce and fashion teams running creative pipelines through an agent.
- ★★ mcp-video-gen is an Apache-2.0 self-hosted server on GitHub, deployed through Portainer, that can expose a local ComfyUI instance and optionally a host-installed Blender. · Blender
- ★★ mcp-video-gen is published behind your own endpoint, typically through a Cloudflare Tunnel with Cloudflare Access verification.
- ★★ Beyond generation, mcp-video-gen offers FFmpeg, OpenTimelineIO timelines, subtitles, whisper.cpp transcription and Piper speech.
- ★★ mcp-video-gen has no built-in generation workflows, memory or skills; its quality is that of your ComfyUI setup.
- ★★ mcp-video-gen is maintained by a single contributor, with the last release two months before Higgsfield's check.
- ★★ Higgsfield's blog calls mcp-video-gen best for developers who want video generation on their own infrastructure.
- ★★ A video render usually takes longer than a single MCP call will wait, so video MCP servers document a job ID plus a status tool.
- ★★★ Higgsfield's blog advises following a long render from an agent with the server's documented status or wait tool rather than a second generate call.
- ★★★ Higgsfield's route depends on the client: the MCP connector in Claude web and Desktop, the official plugin in ChatGPT, the Marketplace integration in Cursor, and CLI + Skills for Claude Code, Codex, OpenClaw and Hermes. · Higgsfield MCP, Claude, ChatGPT, Higgsfield ChatGPT Plugin, Cursor, Higgsfield CLI, Claude Code, Codex, OpenClaw, Hermes Agent
- ★★★ In Claude, Higgsfield's blog points to Higgsfield for multi-model production with characters and audio, HeyGen for presenters and localization, fal.ai for raw models; in ChatGPT, the vendor's official plugin where one exists. · Claude, ChatGPT, Higgsfield MCP
- ★★ In Cursor and Claude Code, Higgsfield's blog says fal.ai gives the most control over models and queues, Vivideo works with an API key, and Higgsfield brings its full catalog. · Cursor, Claude Code, Higgsfield MCP, Higgsfield CLI
- ★★ A clip generated from Claude with Higgsfield MCP shows up in Assets next to higgsfield.ai generations, can reuse a Soul character trained there, and is paid from one balance. · Higgsfield MCP, Claude, Soul ID
- ★★ Three video MCP servers document a cost check as a tool: Dora's list models, fal.ai's get pricing and Vivideo's estimate credits. · credit
- ★★ Higgsfield, Runway, HeyGen, OpenArt and Imagine.art MCP servers sign in through OAuth with no key; Dora and AITuber offer OAuth or an API key for unattended agents. · Higgsfield MCP
- ★★ No video MCP server in the list charges for the connection: Higgsfield, Runway, HeyGen, OpenArt and Imagine.art deduct plan credits, while fal.ai, Vivideo and Dora bill per run. · Higgsfield MCP, credit
- ★★★ Higgsfield MCP and the Higgsfield API are two products with two billing systems; Higgsfield advises MCP for chat and agent workflows and the API for code. · Higgsfield MCP, Higgsfield API
- ★★ Through Higgsfield MCP, generated clips run up to 30 seconds, depending on the model. · Higgsfield MCP
Threads in this video
- restate · Should you use a multi-model platform or a model's own platform? — Higgsfield ranks Higgsfield MCP first because one connector covers a multi-model catalog with characters, audio and ad workflows on one account and one plan-credit balance, with no API key.
- restate · How do you spend fewer credits? — For Higgsfield MCP, Higgsfield advises asking the agent to state the credit cost and wait for your confirmation before generating.
- refine · How do you spend fewer credits? — For video on fal.ai MCP, fal's docs advise using submit job then check job from the start and not resubmitting to check progress, since each resubmission is a new billable job.
- restate · How do you connect a chat or coding agent to a generation platform? — Higgsfield's route depends on the client: the MCP connector in Claude web and Desktop, the official plugin in ChatGPT, the Marketplace integration in Cursor, and CLI + Skills for Claude Code, Codex, OpenClaw and Hermes.
- refine · How do you connect a chat or coding agent to a generation platform? — In Claude, Higgsfield's blog points to Higgsfield for multi-model production with characters and audio, HeyGen for presenters and localization, fal.ai for raw models; in ChatGPT, the vendor's official plugin where one exists.
- first on record · How do you follow a long video render from an AI agent? — Higgsfield's blog advises following a long render from an agent with the server's documented status or wait tool rather than a second generate call.
Mentioned
Central: AITuber · ChatGPT · Claude · Cursor · Dora · fal.ai MCP · HeyGen Remote MCP · Higgsfield MCP · Imagine.art MCP · mcp-video-gen · OpenArt MCP · Runway MCP · Vivideo MCP
Also: Organizations: Facebook · Instagram · Notion · TikTok · Twitter · YouTube
Concepts: AI avatar · credit · lip sync · motion transfer · native audio · Skill · voice cloning
Products: Ad Multiplier · After Effects · Blender · Cinema Studio 4.0 · Claude Code · Claude Desktop · Codex · ComfyUI · Elements · Gen-4.5 · Genjutsu · GPT Image 2 · Grok · Grok Bot · Grok Video 1.5 · Hermes Agent · HeyGen Video Agent · Higgsfield API · Higgsfield ChatGPT Plugin · Higgsfield CLI · Higgsfield Unlimited · Kling 2.6 · Kling 3.0 · Kling 3.0 Omni · Manus · Marketing Studio · MiniMax H3 · Nano Banana Pro · OpenClaw · Photoshop · PixVerse V6 · Seedance 1.5 · Seedance 2.0 · Seedance 2.5 · Soul ID · Veo 3 · Veo 3.1 · Wan 2.7 · Wan 3.0 · Website Building skill
← Previous: One Edit, Three Platforms, Any Language · Next → Aura Edit vs Teaser: What's the Difference?