---

name: vidseeds-projects description: Use for VidSeeds MCP project workflows - create a project from YouTube, list/get/snapshot, regenerate metadata (sync or async), translate metadata, apply thumbnails, and update per-platform config before publish. license: MIT

Projects (MCP)

A project is the workspace for one video across YouTube, TikTok, Instagram, Facebook, LinkedIn, and X.

Discover projects

• vidseeds_list_projects - browse with limits/filters per tool schema.
• vidseeds_get_project - single project record.
• vidseeds_get_project_snapshot - project + platform config in one call (prefer over many gets).

Create from YouTube

1. vidseeds_create_project_from_youtube with the YouTube video id/URL (see tool inputs).
2. If metadata generation is async, poll vidseeds_get_create_project_metadata_job until completed.
3. Confirm with vidseeds_get_project_snapshot.

Regenerate platform metadata

Re-runs optimized titles/descriptions/tags per platform from the stored transcript.

Mode	Tool	Notes
Fast / small	vidseeds_regenerate_project_metadata	Can exceed ~100s edge timeout on large projects
Production	vidseeds_regenerate_project_metadata_async	Returns jobId immediately
Poll	vidseeds_get_regenerate_project_metadata_job	Same result shape as sync when done

Pass focus: "titles" (with regenerate) to use the lightweight titles-only generator (still 1 seed, but far less AI work than full variants).

Voice: If the user has multiple YouTube channels, pass channelId (UC…). With one connection, omit for auto-resolve. voiceBlendRatio optional (tool default documented in tools/list).

Reuse the same requestId on transport retries to avoid double seed charges when the tool allows it.

Keep create, regenerate, and targeted field-refresh calls for the same video close together when possible. VidSeeds.ai reuses recent per-video analysis context for about an hour during a metadata workflow, so follow-up quality checks and regenerations are usually cheaper and faster when they stay in the same project flow.

Targeted / cheap regen (2026-06): Use fields: ["description", "tags"] (or include "title") on regenerate/optimize for light per-field generators (correct lower AI cost for the work, while the action still uses the reoptimize 1-seed price). For thumbnails or edits, use the dedicated thumbnail tools (pricing is per-image with model; edits that reference an input image cost more than pure generation).

Optimize metadata without a full project (standalone)

• vidseeds_optimize_marketing_metadata / vidseeds_optimize_marketing_metadata_async - needs transcript (e.g. from vidseeds_transcribe_audio or YouTube transcript tools).
• Poll: vidseeds_get_optimize_marketing_metadata_job.
• Pass focus: "titles" for title-only (light generator, 1 seed).

Suggest titles only (light 1-seed path for "new titles" requests)

• vidseeds_suggest_titles - dedicated tool for title suggestions only. 1 seed. Accepts projectId (loads transcript/meaning) or direct transcript. Much lower AI cost than full optimize/regenerate. Use this (or regenerate+focus=titles) instead of create_new when user asks for title options. Idempotent with key.

Translate project metadata

• vidseeds_translate_project_metadata - localization across supported languages (no voice dubbing on VidSeeds.ai).

Thumbnail on project

• Generate per vidseeds-thumbnails, then vidseeds_apply_thumbnail_to_project.

Per-platform publish config

• vidseeds_update_project_platform_config - enable platform, title/description/tags, schedule, connection, thumbnail URL override.

Publish

After config is ready, use vidseeds-publishing (vidseeds_preflight_publish → vidseeds_confirm_publish / vidseeds_publish_project).

History

• vidseeds_list_optimization_history, vidseeds_get_optimization_history_item, vidseeds_reoptimize_history_item - past runs and re-run from history.