FrameThrower MCP Server
Enables connecting any MCP client to a cinematography reference library of 5,489 films. Allows searching film frames by look, mood, or craft attributes, and refining through similar-frame discovery.
README
FrameThrower MCP Server
Film references, inside your agent. Connect any MCP client to a cinematography reference library of 5,489 films, indexed frame by frame on lighting, lens character, shot size, colour and mood.
Ask for a look in conversation and the agent comes back with real frames, credited to their film, director and cinematographer.
https://framethrower.ai/api/mcp
Remote server, OAuth 2.1, four tools, no rate limits, $2 of credits free on signup (no card).
Quick start
Claude (Desktop / Web)
Settings → Connectors → Add custom connector → paste:
https://framethrower.ai/api/mcp
Sign in when the browser opens. That's it.
Cursor
One-click:
Or add to ~/.cursor/mcp.json:
{
"mcpServers": {
"framethrower": {
"url": "https://framethrower.ai/api/mcp"
}
}
}
Clients that only speak stdio
Bridge with mcp-remote, which
handles the OAuth flow for you:
{
"mcpServers": {
"framethrower": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://framethrower.ai/api/mcp"]
}
}
}
Full per-client steps in docs/install.md.
The four tools
| Tool | Credits | What it does |
|---|---|---|
search_frames |
2 | Search by concept, mood, colour, composition or scene, in natural language |
find_by_craft |
2 | Find frames by cinematography attribute — lens, shot size, style, setting, time of day, director, year |
find_similar |
2 | More frames like one already shortlisted. This is the refinement loop |
get_frame_details |
2 | Full metadata, thumbnail and link for a single frame |
search_frames
| Argument | Type | Notes |
|---|---|---|
query |
string, required | e.g. "neon-lit rainy street at night" |
limit |
number, 1–40 | defaults to 12 |
find_by_craft
All optional — combine them to narrow.
| Argument | Type | Notes |
|---|---|---|
lens |
string | "anamorphic", "spherical", "vintage_soft" |
shot_size |
string | "closeup", "wide", "medium" |
visual_style |
string | |
setting |
string | "interior" or "exterior" |
time_of_day |
string | "night", "day" |
director |
string | |
year_min / year_max |
number | year range |
limit |
number, 1–40 | defaults to 12 |
find_similar
| Argument | Type | Notes |
|---|---|---|
frameId |
string, required | an id from a previous result |
limit |
number, 1–40 | defaults to 12 |
get_frame_details
| Argument | Type | Notes |
|---|---|---|
frameId |
string, required |
How it's meant to be used
The server sends usage instructions on connect, so a client doesn't need to be told how to behave. The loop it asks for:
- Search — the agent calls
search_frameswith what you described. - Show a few — about four, each with film, director and a link. Not a dump.
- React — "more like #2", "colder", "tighter". That becomes a
find_similaron that frame, or afind_by_craftwith attributes adjusted. - Repeat until it's right.
Every result keeps a link back to FrameThrower, because inline image rendering in chat is unreliable and the gallery is where the frames look like themselves.
See examples/prompts.md for conversations that work.
What it costs
$2 of credits free on signup, no card. Then $1 = 1,000 credits, and each tool call costs 2 — so $1 covers 500 calls. Credits never expire. There are no rate limits.
Run out and the tool replies with a message saying so, including your balance and where to top up, so the agent can explain itself rather than failing silently.
What it returns — and what it doesn't
Tools return metadata, a thumbnail URL, colour palettes and a deep link back to FrameThrower. Never raw image bytes.
This is a discovery and reference tool, not an image-delivery pipe. The frames are not ours to license — they are shown at reduced resolution as references for commentary, education and study, and rights remain with their owners. See intended use.
Why there's no server code here
The FrameThrower MCP server is remote and hosted — you connect to it, you don't run it. So this repo is the connector: manifests, per-client setup, tool signatures and worked examples. Nothing to install, nothing to keep running, and no API key to manage — OAuth means the agent acts as your signed-in FrameThrower account.
Also available
- REST API — 11 endpoints, no rate limits. Machine-readable spec · docs
- npm SDK —
npm install framethrower-ai(source)
Same account, same credits.
About
FrameThrower is the reference app for the future creative: one of the largest film still libraries on the internet, callable from a REST API, an MCP server and an npm SDK. Built by senior filmmakers and ads creatives, for working film professionals.
MIT licensed.
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.