singular-mcp-server
Enables AI agents to read, control, and orchestrate Singular.live live broadcast graphics compositions, including filling control nodes, animating sub-compositions, and running a newsroom rundown-to-graphics workflow.
README
singular-mcp-server
A comprehensive Model Context Protocol server for reading, controlling, and orchestrating Singular.live live broadcast graphics compositions from Claude Desktop or any MCP client.
It lets an agent discover a control app's structure, fill its control nodes (text, images, colors…), animate sub-compositions in and out, push low-latency data, and run a newsroom rundown → graphics workflow over templates you build once in Composer.
What this is (and isn't)
Singular's public API controls compositions; it does not author them. You build the graphics (lower-thirds, full-frames, tickers, bugs) once in the browser Composer, and this server lets agents fill and sequence that kit of parts in real time. Authoring composition structure, account listing, token minting, media upload, and webhooks are not exposed by any public Singular API. See docs/SINGULAR_API.md for the API reference.
Installation
Prerequisites: Node.js ≥ 18 (developed on Node 22) and npm.
# 1. Clone
git clone https://github.com/dqbuilds/singular-mcp-server.git
cd singular-mcp-server
# 2. Install dependencies
npm install
# 3. Build (compiles TypeScript to dist/)
npm run build
# 4. (optional) Verify — runs the mock-API integration test; no Singular account needed
npm test
This produces the runnable server at dist/index.js. Next, configure a Singular
app token and point your MCP client at it (below).
Install globally (optional). From the repo,
npm linkputs asingular-mcp-servercommand on your PATH; then your MCP client can use"command": "singular-mcp-server"instead of an absolutenode dist/index.jspath.
Configuration
Secrets come only from the environment. Copy .env.example → .env (the
server itself reads real environment variables; use your process manager or the
MCP client config to set them).
| Variable | Required | Default | Purpose |
|---|---|---|---|
SINGULAR_APP_TOKEN |
no | — | Default control-app token when a tool omits app/appToken. |
SINGULAR_API_BASE |
no | https://app.singular.live/apiv2/controlapps |
Control REST base. |
SINGULAR_DATASTREAM_BASE |
no | https://datastream.singular.live/datastreams |
Data Stream base. |
SINGULAR_DEFAULT_SUBCOMPOSITION |
no | — | Optional default sub-composition name. |
SINGULAR_REGISTRY_PATH |
no | ~/.singular-mcp/registry.json |
On-disk alias→token registry (owner-only perms). |
SINGULAR_HTTP_TIMEOUT_MS |
no | 30000 |
Per-request timeout. |
LOG_LEVEL |
no | info |
debug|info|warn|error (logs go to stderr). |
Where do tokens come from? Copy each app token from the Singular Dashboard
(inspector "i" → URLs and Token; UNO </>; or Studio → Manage Access → App
Token). The API cannot mint them.
Claude Desktop
Add to claude_desktop_config.json:
{
"mcpServers": {
"singular": {
"command": "node",
"args": ["/absolute/path/to/singular-mcp-server/dist/index.js"],
"env": { "SINGULAR_APP_TOKEN": "your-token-here" }
}
}
}
You can omit SINGULAR_APP_TOKEN and instead register apps by alias at runtime
with singular_register_app.
Tools (25)
Every tool takes response_format (markdown default, or json) and returns
both human text and structuredContent. App-targeting tools accept an app
alias (from the registry) or a raw appToken, falling back to the env default.
Discovery & read
singular_get_app_metadata— ids, name, output/preview URLs.singular_get_model— structural model: sub-compositions + typed control nodes.singular_get_control_state— current live state + values.singular_list_subcompositions— model + live values joined (best overview).singular_find_nodes— filter nodes bytype(e.g.image) / title.
Content control
singular_update_content— fill nodes across one or more sub-comps in one PATCH.singular_set_image— set an image node by URL (validates it's an image node + URL reachability).singular_reset_nodes— reset nodes to their reset values (clear a template).
Animation & playout
singular_animate_state— take sub-comps In/Out (multiple targets per call).singular_update_and_animate— fill + take to air atomically (no stale flash).singular_take_out_all— clear the whole app (TakeOutAllOutput).
Data streams
singular_register_data_stream— store a stream's private token by alias.singular_list_data_streams— list registered stream aliases.singular_push_datastream— PUT low-latency JSON (≤ 60 KB) to a stream.
Media & output
singular_get_output_urls— live output / broadcast / thumbnail URLs.singular_check_image_url— pre-flight an image URL.
Orchestration & rundown
singular_register_app/singular_list_apps/singular_remove_app— the alias registry.singular_map_rundown_template— bind an item type → sub-comp + field map.singular_list_rundown_templates— list mappings.singular_prepare_item— dry-run: resolve a story item to a payload, no air.singular_play_rundown_item— take an item to air (optional server-side auto-out).singular_poll_onair_state— poll what's on air (+ diff since last poll).singular_server_info— server config/health (no secrets).
For agents
The server sends usage guidance to clients as its MCP instructions at connect
time, and every tool has a self-contained description with its args and return
shape. The full playbook — the discover-before-write rule, the canonical flow,
value formats, and gotchas — is in AGENTS.md.
Newsroom workflow example
1. singular_register_app { alias: "evening-news", token: "…" }
2. singular_find_nodes { app: "evening-news", type: "image" } # discover the kit
3. singular_map_rundown_template {
name: "lower-third", appAlias: "evening-news",
subCompositionName: "LowerThird",
fieldMap: { headline: "titleNode", photo: "bannerImage" } }
4. singular_prepare_item { template: "lower-third",
fields: { headline: "Markets rally", photo: "https://cdn/…jpg" } } # verify
5. singular_play_rundown_item { template: "lower-third",
fields: { headline: "Markets rally", photo: "https://cdn/…jpg" },
auto_out_ms: 8000 } # to air
6. singular_poll_onair_state { app: "evening-news" }
An agent can drive this from a rundown/PDF: extract stories + image URLs, map them to templates, and play them to air.
Security
- App tokens are bearer-equivalent secrets. They are stored server-side in the
registry (file mode
0600, dir0700), never returned to the model, and redacted from logs (including inside request URLs). - Host images on your own reachable URL; there is no upload API.
- stdio logging is on stderr only (stdout is the JSON-RPC channel).
Limitations (from Singular's API, not this server)
Authoring compositions, listing an account's assets, minting tokens, uploading media, server-side rendering, and webhooks are all unavailable via public API. On-air awareness is by polling. Auto-out scheduling is in-memory (not restart-safe). See docs/SINGULAR_API.md.
Roadmap
- Optional Streamable HTTP transport.
- Data Node tools over the
apiv1Data Node API. - Batch rundown ingestion helpers (PDF/rundown → items).
- In-memory control-model cache with TTL.
Development
npm run dev # tsx watch
npm run typecheck # tsc --noEmit
npm run inspector # MCP Inspector against the built server
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.