Free Remotion and HyperFrames templates and creative guides; optional hosted Studio workflows.
Scored across 20 tools
Most tools have clearly distinct targets (list vs get vs scaffold vs prepare), and descriptions go out of their way to state boundaries, e.g. prepare_announcement vs prepare_video_template and the two schema tools. The main risk area is the several creation pathways (create_video_brief, edit_video_brief, propose_video_storyboard, prepare_*), which are distinguishable only after reading long descriptions.
Every name follows a consistent snake_case verb_noun pattern with a small, predictable verb set (create, edit, get, list, prepare, propose, review, scaffold, diagnose). Prefixes map cleanly to operation types, so the surface is easy to scan.
20 tools is on the heavy side, but the server genuinely spans two domains (hosted briefs/renders and local VideoSpec/carousel scaffolding) plus discovery metadata. Seven get_* metadata tools add overhead but each covers a different artifact (capabilities, two schemas, guide, example, instructions, template).
The lifecycle is largely covered: discovery, listing, planning/storyboard, preparation, render, status polling, error diagnosis, review, and copy edits via approval. Gaps are modest — no cancel-render or delete/archive-project operation, and publishing is deliberately out of scope, which agents can work around.
Create a private hosted video and start rendering after the user confirms the complete brief and plan usage. Requires account linking. Automatic supplemental footage defaults to off; set automaticBroll to auto only if the user explicitly requests it. Preserve visual exclusions in scenes[].generationContract.prohibitedChanges, separately from the visible copy. Use real source media or explicitly accepted generated visuals for demonstrations; never silently replace requested footage with text. Plan distinct shots for a story, with each generated motion shot at most 15 seconds. For silent videos set narrationMode off and supply explicit scenes; omit script and any unsupplied caption. publishTarget is an optional layout hint, not a publish action; preserve the accepted aspectRatio. Put exact spoken words in script for narration, never filmmaking directions. Supplied originals must be uploaded and bound; do not replace missing branded media with stock. This is the hosted brief contract, not VideoSpec or paired-announcement source. Preserve requestId when retrying the same request, including after a timeout; use a new UUID for changed content. Returns a queued job, not a finished video. Does not publish.
Inputs: title (required), confirmed (required), objective (required), requestId (required), +18 more.
Classify a local validation or render error and return a safe next step.
Inputs: errorText (required).
Prepare exact title, script or caption changes for an existing private video. Returns a Studio approval link; this call does not save copy or start a render. The signed-in user must review and approve there. Approval changes the current run, consumes rerender credits and revokes existing review links. Previous run records are retained. Does not publish. Preserve requestId for unchanged retries.
Inputs: jobId (required), confirmed (required), requestId (required), title, +3 more.
Call first. Reports public planning features and optional account features. Public access metadata does not indicate session authentication; use list_video_projects to verify private read access.
Inputs: no named parameters.
Return the JSON Schema for deterministic Instavar carousel decks.
Inputs: no named parameters.
Read practical guidance before drafting concise plain or poetic copy, choosing carousel stories, designing visuals or recording human feedback. Returns guidance for the calling assistant to apply, not generated copy or saved feedback. Omit guideId to list guides.
Inputs: guideId.
Return a valid starter example for one template family.
Inputs: templateId (required).
Return the clone, install, validate, preview, render, and QA workflow.
Inputs: client.
Get one template's purpose, aspect ratios, scene kinds, and example identifier.
Inputs: templateId (required).
Check the requested private render and its output URL, errors and recovery guidance. Keep workspaceId from creation. When a video is available, call review_video_project to show playback. A queued or rendering state is not completion. Do not automatically submit another render on failure.
Inputs: jobId (required), workspaceId.
Return the published JSON Schema for VideoSpec version 1.0.
Inputs: no named parameters.
List the account-free editorial carousel story structures.
Inputs: no named parameters.
List the published video template families, each available in Remotion and HyperFrames, and optionally filter by output aspect ratio.
Inputs: aspect.
List recent private Instavar video projects after the user links their account. Returns the currently resolved workspace in accountContext; billing details are included only with billing.read permission. Recorded credits are a read-only snapshot, not a reservation or render eligibility guarantee. Use this before review_video_project when the user has not supplied a project ID. Do not infer current workspace or allowance from an older job.
Inputs: limit, workspaceId.
Prepare one paired announcement-card-v1 template, continue its caller-held project, edit copy or switch frameworks. Returns source files for local rendering. For the paired announcement-card-v1 or a HyperFrames announcement request, use prepare_announcement. Never infer a framework from creative wording. Omit framework unless the user explicitly chooses one or says no preference. If needs-selection, ask the user. Continue with the latest returned project and expectedRevision; apply edits through edits before switching. Keep the returned project document for later calls. Never describe this caller-held document as a saved hosted project. VideoSpec templates use prepare_video_template in either framework.
Inputs: copy, edits, project, framework, +1 more.
Validate one VideoSpec against the paired template contract and return local source for the chosen framework, or switch an existing project's framework. For a VideoSpec template in Remotion or HyperFrames, use prepare_video_template with the VideoSpec. Never infer a framework from creative wording; omit framework unless the user chose one or said no preference, and ask the user on needs-selection. Continue with the returned project: edit project.spec, then call again to regenerate or to switch frameworks. A rejected field is unsupported in both frameworks; fix it rather than dropping it. The tool returns local source and commands; it does not render, save or upload.
Inputs: spec, project, framework.
Turn an ordinary video request into a three-beat visual proposal and recommended template. Pass templateId to preserve an accepted template when changing only the aspect; omit it for a new direction. This tool does not generate images, save a project, render, or publish.
Inputs: brief (required), cta, body, title, +3 more.
Open a focused review of one private Instavar video after account linking. This tool does not edit, render, approve, or publish.
Inputs: jobId (required), workspaceId. Returns: jobId, title, status, message, +5 more
Build a deterministic editorial carousel deck from accepted copy. This does not render, save, or publish it.
Inputs: hook (required), payoff (required), points (required), templateId (required), +4 more.
Return a deterministic VideoSpec skeleton. This tool does not generate creative copy or write files.
Inputs: title (required), templateId (required), cta, body, +4 more.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Hosted MCP for Remotion videos: search Remocn components, read docs, plan a video, preview it.
Video production studio for AI agents: AI media, motion graphics as code, timeline and export.
Animated website templates, cinematic video backgrounds, UI components and design systems.
Video API for automated video production at scale: templates, AI voiceovers, subtitles, AI images
Add one secure layer between your agents and this server.