How do I install CueFrame?

Copy the http config for Cursor, Claude Desktop, Claude Code, or VS Code on this page. It uses the publisher's install, not a PlugMCP proxy.

Does CueFrame need an API key?

No required environment variable is listed on this record.

When was CueFrame last checked?

Last checked 2026-10-18. Turn footage into finished video: reframing, captions behind the subject, motion graphics, renders.

CueFrame MCP server

Power your agent

Add this server

The publisher's own URL. Not a PlugMCP endpoint.

Add to Cursor Add in Claude Add to VS Code Run on PlugMCP

Host a copy with our keys and logs.

Server URL

https://api.cueframe.ai/v1/mcp

The Claude button above opens Claude with the name and URL filled in. You still confirm it there.

shClaude Code

1claude mcp add --transport http ai.cueframe-cueframe https://api.cueframe.ai/v1/mcp

Run in a terminal.

jsonClaude Desktop

1{2  "mcpServers": {3    "ai.cueframe-cueframe": {4      "command": "npx",5      "args": [6        "-y",7        "mcp-remote",8        "https://api.cueframe.ai/v1/mcp"9      ]10    }11  }12}

claude_desktop_config.json

Remote · HTTPPublisher installStep-by-step: add this in Claude

Remote URL or a command on your machine? How the two installs differ

Run on PlugMCP

Host a copy with our keys and logs.

A new server you create. It is not this listing's official endpoint. Use the publisher install when you want their URL and their credentials.

Launch

Server facts

Remote

Transport: http
Version: 1.2.0
Last checked

60 tools

Tools

tools/listVoice mode preview

Tool	What it does
compose	Start an async Director compose job: it authors an ensemble of full composition candidates, verifies each, and saves the winner to the project. Body must set EXACTLY ONE source (mutually exclusive): `briefId` — author FROM a persisted Brief (the PLANNED path: create_brief first; every Brief field is
apply_composition	Atomically apply an array of CompositionOps to a project composition. All ops succeed or none do (transactional). Call describe_composition_ops without type for the compact operation list, then with a type for its exact JSON Schema. For negotiated component execution, read get_composition with compo
validate_composition	Dry-run a full Composition object against the SAME pure, DB-free authoring guards saveComposition enforces, WITHOUT persisting it (no project needed). Checks: wire shape (Zod), primitive membership (unknown_primitive / custom_missing_sourcecode), required-param shape (invalid_primitive_params), clip
create_render	Queue an MP4 render of a project's composition. Returns { id (renderId), status, warnings? } — `warnings[]` carries non-fatal advisories (omitted when none). The render themes from the project's brand kit (bind one via create_project / update_project with brandKitId — or PATCH /projects/:id over RES
preview_frame	Queue durable still-frame evidence for layout/copy judgment. Returns immediately with {jobId, kind:'preview_frame', status}; then call wait_job(kind="preview_frame", id=jobId) to receive every presigned frame as an image. The immutable snapshot records geometry, layers, warnings, and whether the tar
preview_component	Queue a typed preview artifact for ONE org-installed component — ASYNC: returns a jobId immediately; then wait_job(kind="preview", id=jobId). The preview and production compositor consume the same content-addressed component bundle and verified artifact contract. `body.componentId` selects the compo
preview_clip	Render a scoped, cheap MOTION preview — ONLY body.fromSec→body.toSec of the composition (max 60s) at preview quality (capped resolution, lighter encode) through the real render pipeline. THE way to judge motion — a cut, a transition, a component's animation — at full fidelity BEFORE paying for a ful
score_composition	Queue the durable server-side judge — it snapshots the exact authored target, samples eval beats, captures frames asynchronously, and grades editorial/spatial/brand/caption. Then wait_job(kind="verify", id=jobId) for per-criterion scores, weighted composite, worst-first critique, and checkpointId. `
consult	Ask the Director for editing advice on the project's CURRENT composition — it NEVER writes your composition (you stay the decider). Async: returns a consultJobId; then wait_job(kind="consult", id=consultJobId, projectId=...) surfaces the advisory envelope (assessment / proposals / principles / plan)
resume	Answer the checkpoint packet wait_job reported as awaiting_input (its {checkpointId} — read the packet FIRST with get_checkpoint and relay the stills + money triple to your human). ONE `body.decision` per gate: 'approve' (proceed; comments allowed — they record as taste without gating), 'revise' (th
create_brief	Persist the video's INTENT as a Brief — the Director's rich input contract (spec: the storyboard skill's output lands here). Fields: goal (required), audience, platform, format, durationSec, tone, cta, locale (output language — all on-screen copy + captions), hook, caption, beats[] (each with an edi
get_brief	Read a brief by id — the persisted intent plus its current quote (compose + estimated render).
list_briefs	List briefs. DEFAULT shows human-authored briefs only; kind:"clip" shows the machine-minted clip briefs (suggest_briefs output), kind:"all" shows everything. Each carries its quote.
get_checkpoint	Read a checkpoint packet — the durable approval unit a held compose pauses at (wait_job reports awaiting_input with {checkpointId, stage, version}). Carries the stage artifact, presigned storyboard stills (REVIEW THE EVIDENCE, not raw JSON), critic verdicts, the money triple {thisStepUsd, jobSpendSo
list_checkpoints	List checkpoint packets newest-first (filter with composeJobId for one run's v1→vN gate history, or projectId). The version stack with per-version status + decisions IS the review history — nothing is deleted.
update_brief	Edit a brief (partial). REFUSED with 409 while an ACTIVE compose job references it — finish, resume, or abandon the job first (the Director never has its input edited mid-run). Edits apply to the NEXT compose.
delete_brief	Delete a brief. Compose jobs already run from it are unaffected (they snapshot).
select_candidate	Select one candidate from a finished compose/consult ensemble as the project's saved composition (whole-candidate replace). Body REQUIRES both fields: { candidateIndex, selectStrategy: "agent-authoritative" } — candidateIndex from wait_job(kind="consult") proposals (mode=amend) or the compose winner
create_component	Create and store a custom component for the tenant: a native scene that arranges its declared images and videos as tiles in 3D space. Returns the ComponentResponse (id + metadata) you can then reference from a composition. No project arg — components are tenant-scoped. SOURCE CONTRACT (version 4; th
list_catalog	Use projection:'component-library' for the unified source-free library page, returned as complete JSON with exact scoped definition/preset/recipe cards and pagination. Choose origin:'builtin' for the maintained builtins or origin:'library' for the authenticated serving home; use libraryKind, categor
resolve_builtin_component	Resolve a selected builtin component-library card using its exact collection, id and revision. Returns the canonical pinned document and source snapshots for a preset or recipe, or the exact source for a definition. Preserve documentSha256 and all identities; a missing revision refuses without falli
apply_template	Insert a resolved builtin component package into an existing project. Pass template as the selected builtin id and body with projectId, the resolved documentSha256, insert:{startTime,idPrefix}, parameter values and explicit media bindings. Read get_composition first and retain expectedRevision, comm
get_component_source	Read an exact component source as complete JSON. Supply sourceVersionId from the selected catalog card to read that immutable revision, retaining sourceSnapshot with its complete versioned definition, manifest and template beside its exact pin and schema metadata. For an explicitly selected builtin,
get_media_context	Fetch bounded authoring context for a media item: dimensions/thumbnail, compact face roster, transcript counts/sample, temporal grids, and fact readiness. The full transcript body is deliberately elided so this read cannot overflow MCP; follow with get_transcript for exact paginated word timings.
get_transcript	Read exact source-time transcript words in bounded utterance pages. Word start/end values are seconds from the source media start. Pass limit (20 recommended), then pagination.nextCursor as after; the formatter omits the unbounded fullText duplicate.
prepare_media	Prepare the perception artifact an authored edit needs BEFORE sighted review. Pass body.kind="subjectTrack" for face/active-speaker reframing, or body.kind="matte" for a true zPlane:"behind-subject" graphic. Pass body.intent {startSec,endSec} using SOURCE time (the media trim, not timeline time); it
get_media_facts	Read prepared subject tracks and behind-subject mattes for one media item. For an exact readiness check, pass kind + startSec + endSec together using the same SOURCE window sent to prepare_media; exact.state is pending \| ready \| failed \| available \| unavailable. If subjectTrack was prepared without
list_media	List the media in your CueFrame library. Filter a real library with kind (video\|audio\|image\|gif\|asset\|source\|stock), source (upload\|generated\|import-url\|stock), or tag — so you find YOUR footage, not e2e debris. Returns media item IDs and metadata; use an ID with get_media_context, suggest_briefs, o
create_project	Create a new project — the unit you compose and render into. compose / apply_composition / create_render / the preview family all require a projectId; this mints one. Pass name (required) plus optionally a format (aspectRatio/fps/resolution) and brandKitId. brandKitId binds a brand kit to the PROJEC
new_composition	Create a NEW composition (a new video) inside a project — a project is a workspace that holds many videos. Returns { compositionId, format, etag }. Pass an optional name and format (aspectRatio/fps/resolution); format defaults to the project's compositionFormat. The first composition becomes the pro
get_composition	READ BACK a composition as JSON — the edit itself: tracks, clips, card HTML, reframe segments, captions, markers, format. This is what create_render encodes, so it is also what you hand to validate_composition (which needs a full inline composition body), what you diff after an edit, and what you ar
list_compositions	List every composition (video) in a project, with its format and which one is active. Use a returned compositionId with create_render(compositionId=...) to render a specific video, or apply_composition to edit it.
delete_composition	Delete a composition (and its version history) from a project. REFUSES the ACTIVE composition (409) — make another one active first, or delete the project. Use list_compositions to see which is active.
list_projects	List your projects — re-orient in an existing workspace instead of minting duplicates (steady-state entry, A9). Each row carries the id every other tool needs.
derive_composition	Derive a reframed SIBLING composition for a different aspect ratio (the agent-native Auto-Reframe / Magic-Resize). Pass projectId + body { fromCompositionId, targetFormat: { aspectRatio, fps?, resolution? } }. CueFrame copies the source composition's authored intent and content-aware-reframes every
import_media	Import a media item into your library from a public URL — the agent-friendly ingest path (single call, no upload protocol). Pass url (https) + filename (+ optional contentType). Returns { id, status:"importing" }. Processing is async: call wait_job(kind="media", id=mediaItemId), or register a webhoo
generate_media	Generate one first-class Library media item from a prompt. Choose generator: text-to-video \| text-to-image \| image-to-video \| text-to-music \| text-to-speech \| text-to-sfx \| auto. Put supported presentation controls in body.style and first/last/image references in body.refs. Returns { id, status:"gen
create_upload	Ingest a LOCAL file — bytes you hold, when there is no public URL to hand import_media (prefer import_media whenever the bytes already sit at an https URL). TWO steps: (1) this tool reserves a media item and returns { id, uploadUrl, expiresAt } — a short-lived PUT URL. PUT the raw file bytes to uplo
finalize_upload	Finalize a direct upload once you have PUT the bytes to the uploadUrl from create_upload. Confirms the bytes landed and starts ingest: image/asset become ready immediately (status "complete"); audio/video enter processing (status "processing"). Returns { id, status, processingPhase }. 400 if no byte
search_resources	Search licensed stock media to compose with across video, photo, sound effects, and GIFs. Pass query for ONE search, or queries[] (≤10) to run them all in a SINGLE call — strongly prefer the batch when you know several things you need. A batch returns results[], one per query in order, each with its
list_render_fonts	The fonts the RENDERER can actually draw — read this BEFORE authoring type, not after a refusal. Per family: the exact family name authoring accepts, the weight RANGES a shipped file declares (a static face has min===max; a variable face gives its axis range), the caption default weights, and whethe
search_fonts	Search web font families by name/style — returns family names to use in composition/component params. Pass query (+ optional perPage ≤50). Read-only + free; surfaces the Google Fonts catalog (family + category + variants + css2 stylesheet URL). IMPORTANT: this searches the GOOGLE catalog, not what t
import_resource	Import a stock candidate (candidateId from search_resources) into a project as an ORG-OWNED media item — provider/externalId/license provenance preserved, bytes deduped across orgs. Pass projectId + candidateId. Returns { id, status:"importing"\|"ready" }; processing is async — call wait_job(kind="me
suggest_briefs	Run AI clip-suggestion analysis on a long-form media item — the CLIPPING path (find the best short-form moments). Returns { status, alreadyRunning? }; then wait_job(kind="suggestions", id=mediaItemId) — its terminal lists each suggestion WITH its briefId. When the media belongs to a PROJECT, each fi
create_webhook	Register a webhook so long-running jobs notify you instead of blocking on wait_job — the right pattern for fan-out/async agents. Pass url + events — ONE terminal event per job kind with status INSIDE the payload (render.completed, compose.completed, media.completed, suggest.completed) plus checkpoin
verify_webhook	Re-run ownership verification on a PENDING webhook — one whose endpoint did not echo the challenge at create time (check its verification.reason: token_not_echoed / non_2xx_<code> / unreachable / ssrf_blocked). After you fix the receiver to reply 2xx echoing data.token, call this to re-send the sign
get_profile	The org's legible taste in ONE read: brand kits, registered critics (the machine-checkable standard both the Director's gates and score_composition enforce), their reference media (presigned thumbnails), and the recent checkpoint-decision log. Read this BEFORE composing for an org you haven't worked
list_critics	List registered critics: criticId, label, tier (rules\|rubric\|evaluator), phase, floor, referenceMediaIds. Rubric text / rules / evaluator URLs are write-only — never echoed. Manage with register_critic (upsert) / delete_critic.
delete_critic	Delete a registered critic by its criticId. It stops running in the Director gates and score_composition immediately.
register_critic	Register a custom eval into the compose critic panel — the critic contract is { rubric (or rules/evaluator), referenceMediaIds?, floor }. Three tiers: rules (declarative constraints — keep-out regions, required elements, max overlays, brand colors), rubric (a text criterion scored by the judge as an
update_project	Update a project's mutable fields (e.g. name, format, brandKitId). Pass projectId + the changed fields as `body`. Setting brandKitId binds/changes the project brand kit so every render of this project themes from it. Returns the updated ProjectResponse.
delete_project	Permanently delete a project you created (cleanup). Removes its composition and render history. Irreversible — only call on a project you own and no longer need.
delete_media	Permanently delete a media item from your library (cleanup). Irreversible. Compositions referencing it will fail to render — only delete media no composition still needs.
update_component	Update an authored component's complete version-4 source/metadata IN PLACE (same id) — the correct verb for the author→preview→fix loop (preview_component returns the admission or runtime error; fix the template here and re-preview, instead of creating a new component each iteration). Pass component
repin_component	Move clips that pin a retired TSX (version 3) component to its native (version 4) successor: the `successorSourceVersionId` that get_composition's `needsUpgrade` (or a `component_source_retired` refusal) names. Pass projectId and, in `body`, optional `clip_ids` (omit them to move every clip that has
update_webhook	Update a webhook subscription (its url or event list). Pass webhookId + changed fields as `body`. The signing secret is NOT rotated here.
delete_webhook	Delete a webhook subscription (stop receiving its deliveries). Irreversible — the signing secret is invalidated.
cancel_render	Cancel an in-flight render (stop a runaway/queued job). Needs projectId + renderId (from create_render). No-op if already terminal.
cancel_compose	Cancel an in-flight compose job (stop a runaway/mis-priced ensemble). Needs projectId + composeJobId (from compose). The director cancel is confirmed before the row flips; any linked charge is released. No-op-reject (422) if already terminal.
cancel_job	Cancel a running async job (verify / preview / brand-kit-extract) returned by an enqueue call — abort a mis-priced job instead of waiting out its TTL. For box-backed jobs the warm box is reaped best-effort; cancel never meters. No-op-reject (422) if already terminal.

CueFrame — MCP, CLI and video API

CueFrame is a hosted service for composing, editing and rendering video. Connect your agent over MCP, automate an edit from the CLI, or integrate the REST API into your application. Each surface works with editable projects containing clips, captions and motion graphics, and renders them to MP4.

Your workflow	Start here
Ask an agent to make or edit a video	MCP and agent plugin
Run video workflows from a terminal or script	CLI
Build video editing into an application	API and TypeScript SDK

First render · Connection guide · Templates · API reference · Agent documentation index

A frame from the Yosemite Peregrines template. Watch the video and try the same edit.

MCP and agent plugin

For Claude Code, Codex, Cursor, VS Code, and other supported local agents, run this in your terminal with Node.js 24 or newer:

npx -y [email protected] install

The installer detects supported agents and adds the hosted CueFrame MCP endpoint to their configuration. It installs the skills in ~/.agents/skills and ~/.claude/skills, preserving locally modified skills by default. It can update multiple detected clients; use --dry-run to inspect the changes first.

Restart your agent session. Then ask:

Use CueFrame's get_account tool to check my connection and show my available credits.

Your client may prompt you to authenticate or open a browser. Sign in to CueFrame and authorize the connection; interactive OAuth clients need no API key. Cloud operations use credits. Check your balance and estimated costs before starting paid work; see cloud rates.

ChatGPT and Claude Desktop: use your client's hosted connector setup with https://api.cueframe.ai/v1/mcp, then authenticate. The terminal installer does not configure ChatGPT. See the connection guide and CueFrame connection skill for supported paths. Skills installation and hosted connector setup are separate steps.

Already connected? Skip installation and give your agent a video task. If CueFrame's tools do not appear, restart the client and check its MCP connection settings.

Make your first video

Try the Yosemite first-render walkthrough: render a template, change one word, and render again from the same project.

Download the full original footage from the National Park Service as yosemite-peregrines.mp4. Use the original source, because the finished demo has its title baked in. Follow the source's download and reuse guidance.
Give your connected agent the source file and this instruction:

Use CueFrame to start the yosemite-peregrines template from cueframe-ai/cueframe-templates with this original video. Load template.cueframe and its prepared subject matte. Preview the editable project and estimate cloud costs. After I approve the render, return the MP4 link and project ID.
After the first render, ask:

In that same project, replace the behind-subject title PHENOMENAL with EXTRAORDINARY. Preserve the footage, spoken captions, audio, timing, placement, and depth. Preview the changed title, estimate the render cost, and render after I approve.

The template selects 83.616–93.359 seconds of the original video. You should get two MP4s from one editable project. Uploading footage and running cloud operations use credits; the two renders are billed separately.

The project stays editable: the desktop editor shows the title as a separate layer, alongside footage and captions. Watch the title change.

What happens over MCP?

flowchart LR
    A[Your request + footage] --> B[Your agent + CueFrame skills]
    B --> C[Hosted CueFrame MCP]
    C --> D[Editable project + preview]
    D --> E[Render job → MP4 link]

The agent inspects available tools, checks your account, prepares media and the project, previews changes, and submits a render. Long-running operations return IDs; the agent waits for completion and gives you the result. You do not need to write the tool calls yourself. The detailed quickstart explains the individual calls.

For your next video, describe the result and provide the inputs:

Turn this interview into a 20-second vertical clip with word-timed captions. Preview the strongest moment and show estimated costs before rendering.

Find an image-carousel component for these product photos. Show suitable choices and their required inputs before creating the video.

Change only the final title in this project. Keep the rest of the edit unchanged.

Other installation options

Choose one setup path. A plugin installs the skills and hosted MCP configuration together. A server-only connection adds tools; skills provide the workflow guidance.

Claude Code:

claude plugin marketplace add cueframe-ai/cueframe-mcp
claude plugin install cueframe@cueframe

Codex:

codex plugin marketplace add cueframe-ai/cueframe-mcp
codex plugin add cueframe@cueframe

Restart the session, authenticate when prompted, and ask the agent to call get_account.

Gemini CLI:

gemini extensions install https://github.com/cueframe-ai/cueframe-mcp

Install skills in any client that supports Agent Skills:

npx skills add cueframe-ai/cueframe-mcp

Skills alone do not connect the MCP server. Add the hosted endpoint through your client's MCP settings if it is not already connected.

The hosted endpoint is https://api.cueframe.ai/v1/mcp, using Streamable HTTP. For example, in Claude Code:

claude mcp add --transport http cueframe https://api.cueframe.ai/v1/mcp

For Cursor and VS Code, use the one-click links in the connection guide. Configuration keys differ between clients. Per-client examples are in examples/.

Interactive clients can use browser sign-in. Headless clients send Authorization: Bearer cf_live_…; see llms-install.md for authentication and stdio bridge options. The glama/ shim executes no tools and cannot be used to make videos.

Skills included

Skills teach the agent how to carry out video tasks. You describe the task; the agent selects the relevant guidance.

Task	Skills
Plan and compose a new video	cueframe-storyboard, composing-video
Clip a podcast or interview	clip-a-talking-head
Make a launch film or social reel	launch-video, make-a-social-reel
Make a brand reel or change an existing brand	brand-reel, rebrand-a-video
Extract a visual style	extracting-brand-kits
Create a product demo	cueframe-brand-demo, cueframe-product-video
Customize motion graphics or make a 3D shot	cueframe-component-authoring, cueframe-scene-shot
Add music or adapt to more formats	add-music-bed, every-format-from-one-edit
Preview, refine, and check quality

Claude Code's session hook and Cursor's rule help select the entry point. Other hosts load the installed skills through their own mechanisms. Scoped edits should preserve unrelated work.

CLI

Use the CLI for explicit video commands, shell scripts and structured JSON output. It connects to the hosted API; it does not run a local renderer. Requires Node.js 24 or newer.

npx -y [email protected] --help
npx -y [email protected] login
npx -y [email protected] api GET /v1/me --json
npx -y [email protected] describe --json

login opens the browser to authenticate. Headless scripts can supply CUEFRAME_API_KEY. The account request checks your connection and credits; describe lists command arguments and options without making an API call.

CLI guide · CLI walkthrough · Templates · npm package

API and TypeScript SDK

Use the REST API to create projects, import media, change compositions and follow render jobs from your application. Requests go to https://api.cueframe.ai with an API key from your account. Check your account before metered operations; cloud rendering and processing use credits.

For TypeScript or JavaScript on Node.js 24+, install the typed client:

npm install @cueframe/sdk

import { account, createClient } from "@cueframe/sdk";

if (!process.env.CUEFRAME_API_KEY) throw new Error("Set CUEFRAME_API_KEY");
createClient({
  baseUrl: "https://api.cueframe.ai",
  apiKey: process.env.CUEFRAME_API_KEY,
});
const result = await account.getAccount();
console.log(result.plan, result.entitlements);

Other languages can call REST directly or generate a client from the versioned public OpenAPI contract. The SDK is a client for the hosted service, not the server implementation.

API walkthrough · API reference · SDK guide · Render example · npm package

What this repository contains

This repository contains the agent plugin, developer toolkit, and registry files. The hosted MCP server is maintained separately in the application repository. You connect to CueFrame's hosted service; cloning this repository does not run that service locally.

Area	What you can use or contribute
`skills/`, host manifests, `hooks/`, `rules/`	Agent workflows, plugin installation, and task routing
`packages/`	The public API contract, TypeScript SDK, CLI, and packaged skills
`examples/`, `llms-install.md`	Client setup and an end-to-end SDK render example
`server.json`, `glama/`	Hosted-server registration and a discovery-only shim
`tools/`, `scripts/`	Generators, validation, packaging, and repository checks

Contribute

Documentation, client examples, skill improvements, and toolkit fixes are welcome. Start with good first issues or read CONTRIBUTING.md for setup, generated-file rules, tests, and commit sign-off.

Edit skills in root skills/; change generated API clients through their documented inputs. Report hosted-service problems in this repository's issue tracker with reproduction steps and a request ID when available. Report vulnerabilities privately using SECURITY.md. Community participation follows the Code of Conduct.

Published packages

Package	Purpose	npm
`@cueframe/api-contract`	The public OpenAPI document, its reviewed operation inventory and a content-addressed manifest: the only input the generated clients read.
`@cueframe/sdk`	Typed fetch-based TypeScript client with zod request schemas and SSE helpers for render and compose progress.
`cueframe`	The CLI: login, projects, media, compositions, renders, exports and `install`, every command with `--json`.
`@cueframe/skills`	The `skills/` tree as an npm package (`SKILLS_DIR` from `@cueframe/skills/dir`), for programs that copy or embed the skills.

The MCP server itself (@cueframe/mcp) is a private package maintained in the application repository; this repository ships the contract it is generated from, the clients, the plugin, the registry entry (server.json) and a discovery shim, not the server.

How the contract flows

private API ──sync──▶ api/openapi.json (+ scene-schema.json, scene-examples.json, contract-manifest.json)
                          │
                          ├──▶ packages/api-contract/dist  ──publish──▶  @cueframe/[email protected] ──▶ the MCP server's generators, third parties
                          └──▶ packages/sdk/src/generated        (Orval: operations, models, zod)

The private API publishes its OpenAPI projection into api/; pnpm generate:api-consumers regenerates the SDK from it. This is the one copy of the contract: the MCP server's generators in the application repository and every other consumer read it from the @cueframe/api-contract npm package, whose version moves with the contract (pnpm check:contract-version fails an api/*.json change without a bump), so the tool surface and the SDK always describe one contract. Nothing under api/ or a generated/ directory is hand-edited: pnpm check:api-inputs proves the three api/*.json inputs came from one sync, and pnpm check:api-consumer-drift proves the generated trees match them byte for byte.

How the skills flow

Root skills/ is the single source: the plugin hosts read it from this repository, and pnpm build copies it into packages/skills/dist (published as @cueframe/skills) and from there into the CLI's dist/skills for cueframe install. pnpm check:plugin validates every skill and manifest; agents/openai.yaml and assets/icon.svg beside each skill are generated.

Registry files

server.json: the MCP registry entry for the hosted server (ai.cueframe/cueframe, remotes only). pnpm check:server-json validates it against the vendored registry schema; the registry job of the publish workflow publishes it.
glama.json: the Glama listing's maintainer file.
glama/ and the Dockerfile: a stdio discovery shim that replays a snapshot of the hosted server's tool list and instructions for registries that introspect over stdio. It is not a server (see below).

Hosted, desktop-local, and the discovery shim

The hosted service, desktop integration, and registry shim serve different purposes.

The hosted server at https://api.cueframe.ai/v1/mcp is the product. Every install line above talks to it; it authenticates with OAuth or an API key and does the composing and rendering. Its package, @cueframe/mcp, is private.
The CueFrame desktop app ships its own local MCP server for the project you have open. It needs no account and is connected through the connection file the app publishes; the app's Agents settings show the exact command, and https://docs.cueframe.ai explains it. Its bundled binary is called cueframe-mcp; it lives inside the app bundle.
This repository's discovery shim (glama/server.mjs, built by the Dockerfile) is not a server. It replays a snapshot of the hosted server's tool list and instructions over stdio so registries that build an image can list the tools without an account; every tools/call answers with a pointer to the hosted server. glama/refresh-snapshot.md says where the snapshot comes from.

Repository layout

skills/<name>/SKILL.md: the skill. agents/openai.yaml and assets/icon.svg beside it are generated (pnpm build:plugin).
.claude-plugin/, hooks/, .mcp.json: Claude Code plugin, session router and server.
.codex-plugin/, .agents/plugins/, mcp.json: Codex and ChatGPT plugin and marketplace.
.cursor-plugin/, rules/: Cursor plugin and the always-on rule.
gemini-extension.json, skills.sh.json, plugin.json: Gemini CLI, skills.sh, the portable manifest (agent-plugins.org).
api/, packages/, tools/, scripts/: the contract, the four npm packages, their generators and checks (scripts/plugin/ holds the plugin's).
server.json, glama.json, glama/, Dockerfile: the registry files.
catalog/: the tool reference and error codes exported from the MCP catalog in cueframe-ai/cueframe (pnpm --filter @cueframe/mcp export:tool-reference -- <this repository>/catalog there). skills/cueframe-connect/references/tools.md is rendered from it, and the skills check reads it: a skill may name only a tool it lists.
examples/, llms-install.md: per-client setup and the agent-readable install guide.

Development

pnpm install
pnpm check      # input provenance, the contract version rule, the public-contract gate, server.json, the plugin, and tsc --noEmit in sdk and cli
pnpm build      # stamps the plugin manifests and skill cards, then tsup in every package; the CLI packages the skills
pnpm test

Node 24 or newer and pnpm (see .nvmrc and packageManager). CONTRIBUTING.md explains which paths are generated, the skill contract and how a change travels; AGENTS.md is the same for an agent working in this tree; llms-install.md connects a client step by step.

License

Apache-2.0. See LICENSE and NOTICE.

CueFrame

Add this server

Host a copy with our keys and logs.

Server facts

Tools

Questions

How do I install CueFrame?

Does CueFrame need an API key?

What transport does CueFrame use?

When was CueFrame last checked?

Documentation

CueFrame — MCP, CLI and video API

MCP and agent plugin

Make your first video

What happens over MCP?

Other installation options

Skills included

CLI

API and TypeScript SDK

What this repository contains

Contribute

Published packages

How the contract flows

How the skills flow

Registry files

Hosted, desktop-local, and the discovery shim

Repository layout

Development

License