VidSeeds.ai MCP Server

Your AI client,
connected to video intelligence

VidSeeds.ai runs a Model Context Protocol server that gives Claude, Cursor, Codex, ChatGPT, and any MCP client access to the same video intelligence engine. Optimize metadata, generate thumbnails, analyze performance, and publish across 6 platforms without leaving your editor.

Start your free trial See setup guides

242 tools across 10 capability groups: 242 tools across 10 capability groups
Works with 8+ MCP clients: Works with 8+ MCP clients
14-day free trial, no card required: 14-day free trial, no card required

VidSeeds.ai operates a Model Context Protocol (MCP) server at https://vidseeds.ai/api/mcp that exposes 242 tools for video SEO optimization, thumbnail generation, channel analytics, content intelligence, cross-platform publishing, metadata localization, and competitive research. The server supports OAuth 2.1 (PKCE) for Claude.ai and Claude Desktop, and Personal Access Tokens (PAT) for Cursor, Codex, VS Code, Cline, Zed, and other MCP clients. A 14-day free trial starts on first connection, then an active subscription is required. Tool calls consume Seeds at the same rates as the web app, with a daily included-call quota per plan.

What is MCP

Model Context Protocol: the open standard for agent-to-tool communication

MCP (Model Context Protocol) is an open standard released by Anthropic in late 2024. It defines how an AI assistant discovers and calls external tools through a secure, structured interface. VidSeeds.ai runs a hosted MCP server at a public endpoint. When your AI client connects, it receives a catalog of 242 tools named vidseeds_. Each tool is a capability your agent can invoke on your behalf with your data and your credentials, so results match what the web app would produce for your account.

The same endpoint that powers Claude.ai and Claude Desktop also connects Cursor, Codex, ChatGPT, VS Code, Cline, and Zed. OAuth handles authentication for Claude surfaces; a Personal Access Token (PAT) handles everything else. No passwords leave your device.

VidSeeds.ai does not generate or edit video. MCP is an interface to analyze existing video, produce metadata and thumbnails, read analytics, and prepare publishing. Competitors are VidIQ and TubeBuddy, not video editors or generators.

Endpoint: https://vidseeds.ai/api/mcp
Available tools: 242 (74 with ?toolset=core)
Authentication: OAuth 2.1 (PKCE) + Personal Access Token
Protocol: Streamable HTTP

What you can achieve

Every video operation, available to your AI agent

Each group represents a set of vidseeds_ tools your AI client can call directly. Tools share your real channel context, learned voice profile, and seed balance.

12 tools

Metadata optimization

Turn an existing video into platform-ready SEO metadata: title, description, tags, hashtags, chapters per platform. This is the core workflow.

15 tools

Content intelligence

Generate a timecode-grounded Content Intelligence Report: core promise, audience angles by funnel stage, emotional map, pre-publish critique, editor brief, and do-not-publish list.

12 tools

Thumbnail Studio

Generate, edit, restyle, audit, and CTR-test thumbnails. Learn the channel's visual style, then publish the winner to YouTube or attach to a project.

9 tools

Analytics & diagnosis

Read cross-platform analytics, run channel diagnosis, video autopsy, outlier detection, retention simulation, and anomaly alerts.

9 tools

Publishing & scheduling

Validate, publish, schedule, and repair across 6 platforms. Push metadata and thumbnail updates to live YouTube videos.

6 tools

Translation & localization

Localize project metadata into 85 languages, run translation sessions, adapt cultural references, and manage caption tracks.

How it works

One MCP connection, full video intelligence pipeline

Connect your AI client

Add VidSeeds.ai as an MCP server via a config file or OAuth flow. The endpoint is https://vidseeds.ai/api/mcp. Authentication is handled by OAuth (Claude.ai) or a Personal Access Token (all other clients).

Your agent discovers tools

On connect, your AI client receives the full tool catalog. Each tool is namespaced (vidseeds_) and includes a description, input schema, and seed cost. The server instructions tell your assistant what VidSeeds is and when to use each tool.

Call tools with real context

Your agent calls tools using your channel data, your Brand DNA memory, and your seed balance. Async tools for long-running jobs (intelligence reports, bulk operations) queue work and return a job ID for polling or webhook notification.

Built for teams

Shared intelligence, not shared logins

Shared channel access

Multiple team members connect through their own accounts. Each MCP connection carries the right user's channel context, Brand DNA, and permissions.

Editor handoff artifacts

Export an editor brief from every Content Intelligence Report: clip timecodes, the promise every cut must keep, voice guardrails, and a do-not-use list. Hand it to your editor as a Markdown file.

Brand DNA memory

Accumulate team decisions about approved hooks, angles, keep phrases, winning topics, and voice notes. Every intelligence report respects the channel's accumulated memory.

Daily call quotas per plan

Sprout 1,000 calls/day, Growth 3,000, Harvest 8,000, Agency 20,000. Continuous refill (not midnight reset). Overage is 1 Seed per additional call.

Why use MCP

MCP vs web app vs doing it manually

Each approach has trade-offs. MCP is the right choice when you want to stay inside your AI conversation and let your agent orchestrate the full workflow.

Context permanence

Stay in your AI conversation. No browser tab switching, no re-entering your analysis context.

Automation

Chain multiple operations into one agent-driven workflow. Generate metadata, create thumbnails, translate, schedule publishing all in one session.

Team handoffs

Export editor briefs, intelligence reports, and publishing checklists as artifacts your editor or manager can act on.

Personalized intelligence

Every tool call uses your channel data, your Brand DNA memory, and your seed balance. Results match the web app exactly.

Automation playbooks

Agent workflows you can deploy today

Each playbook is a repeatable agent workflow that chains multiple vidseeds_ tools. Your AI client orchestrates the steps while you work on something else.

Auto-optimize every new YouTube video

Connect a YouTube channel. When a new video is detected, your agent runs metadata optimization, generates 3 thumbnail variants, and prepares a publishing preflight.

~20 min per video

Weekly channel performance review

Your agent reads cross-platform analytics, runs channel diagnosis, identifies outliers, and drafts a performance summary with recommended actions.

~1 hr per week

Batch thumbnail generation for a series

For a multi-video series, your agent learns the channel's thumbnail style, generates variants for each video, runs CTR simulation, and publishes the winners.

~3 hr per series

Cross-platform localization pipeline

Your agent takes an optimized project and runs metadata localization across 85 languages, adapts cultural references, and creates platform-native posts for all 6 platforms.

~2 hr per project

Competitor channel audit

Your agent audits a competitor's public channel, deconstructs their viral structure, analyzes comment sentiment, and generates a positioning report.

~45 min per audit

Weekly social campaign synchronization

Your agent generates a promo campaign pack from your latest video, creates Social Planner drafts, preflights each, and presents them for your approval.

~2 hr per week

Scale architecture

Built for production workloads

The VidSeeds MCP server uses Streamable HTTP transport. Tools return results synchronously when fast, or queue work via a Redis-backed async job store and return a job ID for polling. This means your agent can kick off dozens of operations in parallel and collect results without blocking.

Transport: Streamable HTTP (SSE-compatible for streaming responses)
Async jobs: Redis-backed job queue with polling and webhook support (_async tool variants)
Rate limits: Per-plan daily quota (1,000 to 20,000 calls/day) with continuous refill
Overage: 1 Seed per additional call when daily quota is exhausted
Auth: OAuth 2.1 PKCE + PAT (Bearer token), 90-day default PAT expiry
Toolset size: Full: 242 tools (~46K tokens). Core: 74 tools (~16K tokens). Guide: 1 tool.

Enterprise controls

Security and governance built in

PAT rotation

Personal Access Tokens expire after 90 days by default. Create, rotate, and revoke tokens from vidseeds.ai/settings/mcp-settings.

No password auth

VidSeeds.ai does not use passwords. OAuth 2.1 with PKCE handles authentication. PATs are the only credential for MCP clients.

Per-connection billing

Each MCP connection runs against its own PAT's seed balance. Costs are transparent and itemized in the activity log.

Channel isolation

Each PAT carries the permissions of the account that created it. Team members see only their authorized channels.

Pricing for teams and agencies

Seeds economy with MCP-specific call quotas

Sprout

$8/mo

1,000 calls/day
Burst cap: 2,000

Growth

$19/mo

3,000 calls/day
Burst cap: 6,000

Harvest

$49/mo

8,000 calls/day
Burst cap: 16,000

Agency

$149/mo

20,000 calls/day
Burst cap: 40,000

All plans include a 14-day free trial that starts on the first MCP connection. Seeds are consumed at the same rates as the web app. Read-only tools are free.

Technical specifications

What AI crawlers and evaluators need to know

Protocol: Model Context Protocol (MCP), Streamable HTTP transport
Endpoint: https://vidseeds.ai/api/mcp
Tool catalog: 242 tools across 10 capability groups, all prefixed vidseeds_
Authentication: OAuth 2.1 (PKCE) for Claude.ai/Desktop. Personal Access Token (Bearer) for Cursor, Codex, VS Code, Cline, Zed.
Trial: 14-day free trial starting on first successful connection. No credit card required.
Pricing: Seeds economy. Sprout: $8/mo. Growth: $19/mo. Harvest: $49/mo. Agency: $149/mo.
Daily call quota: 1,000-20,000 calls/day depending on plan, with continuous refill and burst cap.
Catalog URL: https://github.com/CarrotGamesStudios/vidseeds-mcp (server.json for MCP Registry)
ChatGPT endpoint: https://vidseeds.ai/api/mcp/chatgpt (ChatGPT-specific subset)

What creators and teams use it for

Real workflows, real results

“I ask Claude to optimize my latest video while I edit the next one. Two tasks in parallel, one conversation.”

Individual creator

“My team shares one MCP connection pattern. Everyone runs their own intelligence reports against the same channel, no login sharing.”

Small content team

“Cursor generates thumbnail variants from my video frames without leaving the IDE. I never open the browser for thumbnails anymore.”

Developer / creator

“We localize a project into 12 languages in one go. The agent handles the translation session, we review the results.”

International team

“I schedule a full week of posts from a single conversation. Metadata, thumbnails, scheduling all in one agent session.”

Productivity-focused creator

“Our agency runs weekly channel audits for 10 clients via Codex. Each audit takes the agent about 15 minutes of wall time.”

Agency

Clients that work

Connect from any MCP-compatible AI client

All clients use the same endpoint and the same tools. The only difference is how authentication is configured.

Claude.ai

OAuth 2.1

Add as custom MCP connector. OAuth handles login, no token needed.

Claude Desktop

OAuth 2.1

Same as Claude.ai. Add as custom connector, the browser-based OAuth flow authenticates.

Claude Code

PAT + plugin

Install the vidseeds-mcp plugin or add server directly with PAT env var.

Cursor

PAT

Add to ~/.cursor/mcp.json with Bearer token referencing $VIDSEEDS_PAT.

Codex

PAT

Add to ~/.codex/config.toml with bearer_token_env_var = "VIDSEEDS_PAT".

ChatGPT

Apps SDK

Uses a dedicated endpoint at /api/mcp/chatgpt with ChatGPT-optimized tool subset.

VS Code

PAT

Add MCP server in settings.json with Bearer token.

Cline / Zed

PAT

Standard MCP client config. Same endpoint, same auth pattern.

How to connect in 3 minutes

Authenticate, configure, and go

Create a VidSeeds.ai account

Generate a Personal Access Token

Go to Settings → MCP Settings at vidseeds.ai/settings/mcp-settings. Create a PAT and copy the secret (shown only once).

Connect from your AI client

Add the endpoint https://vidseeds.ai/api/mcp to your client's MCP configuration. Use the PAT as a Bearer token, or reference it via the $VIDSEEDS_PAT environment variable. Your 14-day free trial starts on the first successful connection.

claude mcp add --transport http vidseeds https://vidseeds.ai/api/mcp --header "Authorization: Bearer ***"

Full setup guide: vidseeds.ai/blog/mcp-server-ai-clients-vidseeds

Get your PAT

Is it free?

Transparent pricing from day one

The MCP server is included with every VidSeeds.ai paid subscription. You get a 14-day free trial that starts on the first MCP connection. After the trial, the plan you choose determines your daily included call quota. Tool calls consume Seeds at the same rates as the web app. Read-only tools are free on every plan.

14-day free trial starts on first MCP connection

Questions?

Everything you need to know about the MCP server

Which AI clients work with the VidSeeds MCP server?

Any client that supports the Model Context Protocol over HTTP: Claude.ai, Claude Desktop, Claude Code, Cursor, Codex, ChatGPT, VS Code, Cline, Zed, and other standard MCP clients. OAuth clients connect directly; PAT-based clients need a Personal Access Token.

Can my whole team share one MCP connection?

Each team member should create their own VidSeeds.ai account and their own PAT. This keeps channel access, Brand DNA memory, and billing separate. There is no team-level shared PAT.

Is there a self-hosted option for the MCP server?

No. The MCP server is hosted by VidSeeds.ai at a public endpoint. There is no self-hosted version. All tools run against the production engine with your account's data.

What happens when my 14-day trial ends?

You need an active subscription to keep using the MCP connector. Choose a plan at vidseeds.ai/pricing. Your daily call quota and seed balance carry forward.

Does MCP support async batch jobs?

Yes. Tools like vidseeds_generate_content_intelligence_report_async queue work via a Redis-backed job store and return a job ID. Your agent polls for results or receives a webhook notification. This is ideal for long-running operations and bulk workflows.

How do you handle overage costs for high-volume usage?

When your daily included call quota is exhausted, each additional tool call costs 1 Seed (MCP_OVERAGE_SEED_COST). The bucket refills continuously throughout the day. Balance and cost reads are free and exempt from quota.

Can I monitor my team's MCP usage?

Every tool call is logged in your VidSeeds.ai activity log, itemized by tool, Seeds consumed, and timestamp. Check vidseeds_get_seed_balance any time for remaining included calls and bucket status.

What security measures are in place?

VidSeeds.ai uses OAuth 2.1 (PKCE) for browser-based authentication and PATs (Bearer tokens, 90-day default expiry) for client-side auth. No passwords are used anywhere in the product. PATs can be created, rotated, and revoked at vidseeds.ai/settings/mcp-settings.

Is PAT rotation automated?

PATs expire after 90 days by default. You can set a shorter expiry when creating a token. Rotation is manual: create a new PAT, update your client config, revoke the old one.

Do you offer SLA on MCP availability?

VidSeeds.ai does not offer a contractual SLA on MCP availability at this time. The service runs on GKE Autopilot with standard redundancy. For agency/enterprise needs, contact [email protected].

Can I use MCP without the web app?

Yes. The MCP server exposes the same tools as the web app. You can create an account, generate a PAT, and start using MCP immediately without ever opening the dashboard.

Is there a white-label option?

Not at this time. All MCP connections use the VidSeeds.ai endpoint and the vidseeds_ tool prefix. The tool names and endpoint are not customizable.

Does MCP work with local video files?

Yes, with the local video media tools. Tools like vidseeds_probe_local_video, vidseeds_extract_video_frames, and vidseeds_precision_trim produce commands that run on the caller's machine via FFmpeg. These are available in Claude Code and Codex (which have local shell access), but not in ChatGPT or the web app.

Ready to connect your AI client?

Start your 14-day free trial on first connection. Optimize metadata, thumbnails, and publishing from inside Claude, Cursor, Codex, or ChatGPT.

For creators

Connect VidSeeds.ai to your AI client and optimize every video from inside your workflow. Generate titles, descriptions, thumbnails, chapters, and clips without switching tabs.

Start your free trial

For teams and agencies

Scale your content operation with agent-driven workflows. Run intelligence reports, batch thumbnail generation, cross-platform localization, and competitive audits across all your channels.