About Infiknit
Infiknit is a local MCP server for agents using the Infiknit desktop app to create images, videos, and connected canvas workflows. It is intended for local agents on the same computer, with results available for inspection in the agent.
Key features
- 01Generate images and videos in project folders
- 02Create and connect canvas nodes
- 03Validate nodes before queueing
- 04Inspect results from the agent
Use cases
- 01Generate project assets with Infiknit’s saved keys
- 02Build multi-shot ad or product-shot canvases
- 03Review generated results without leaving an agent
MCP support
The server connects local agents to the Infiknit desktop app running on the same computer. It exposes image and video generation, canvas workflow operations, and result inspection; cloud agents cannot reach its loopback connection.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
We haven’t collected a package or remote endpoint for this server. Follow the instructions in the project README.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project READMEInfiknit for agents
Give Claude Code, Codex, Cursor, Grok Build, and other local agents Infiknit: generate images and videos into a project, or build connected canvas workflows while you watch.
Infiknit is a desktop app. This plugin connects to the copy running on this computer. Cloud agents cannot reach it.
Before you install
- Download Infiknit for macOS or Windows.
- Open Infiknit and keep it running.
- Optional: Settings → MCP Connect already one-click installs the same connection. Use this plugin when you want Infiknit to show up in an agent marketplace.
The launcher reads Infiknit’s local token from the desktop app. It never asks for provider API keys.
Install
Claude Code
/plugin marketplace add Synthenova/infiknit-mcp
/plugin install infiknit@infiknit
Codex
codex plugin marketplace add Synthenova/infiknit-mcp
codex plugin add infiknit@infiknit
Grok Build
grok plugin marketplace add Synthenova/infiknit-mcp
grok plugin install infiknit --trust
Cursor, Copilot, Windsurf, Cline
Add this local stdio server (the launcher finds Infiknit on the same machine):
{
"mcpServers": {
"Infiknit": {
"command": "node",
"args": ["/absolute/path/to/infiknit-mcp/plugins/infiknit/bin/infiknit-mcp.cjs"]
}
}
}
Or clone this repo and point args at plugins/infiknit/bin/infiknit-mcp.cjs. Reload the agent after installing.
What it does
- Generate an image or video into the agent’s project folder using keys already saved in Infiknit.
- Create, connect, and queue canvas nodes for ads, product shots, and multi-shot workflows: one all-or-nothing
applyCanvasTransactionper build,verifyNodesbefore queueing, exact modelcontrolsfromgetCapabilities. - Inspect results without leaving the agent.
The plugin ships the same operating card and full guide the desktop app installs (skills/infiknit/SKILL.md and GUIDE.md), so an agent that installs from this marketplace reads the current tool contract.
Tested workflow resource
For a reproducible creative example with sample inputs, a versioned blueprint, MCP steps, expected outputs, and local-connection limits, see examples/three-shot-ugc-product-ad.md.
Keep Infiknit open while work runs. Replacing the token in Settings → MCP Connect refreshes one-click agent installs; this plugin rereads the new token on the next launch.
Security
- Talks only to
http://127.0.0.1:31415/mcp(or31416for Infiknit Dev). - Uses the local bearer token stored by Infiknit. The token never leaves this computer.
- Does not read provider keys,
.envfiles, or SSH credentials. - Cloud ChatGPT, Claude.ai custom connectors, and Cursor Cloud Agents cannot use this loopback server.
Privacy: https://infiknit.app/privacy
Terms: https://infiknit.app/terms
Manual env override
If you copied MCP config from Infiknit, you can skip discovery:
export INFIKNIT_MCP_URL=http://127.0.0.1:31415/mcp
export INFIKNIT_MCP_TOKEN=your-local-token
Set both or neither.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.