Roboflow Agent Plugin
The Roboflow plugin gives an agent computer vision skills: data, training, model evaluation, inference and platform API calls.
- 54 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- The MCP server works with a Roboflow account and manages projects, datasets and training.
- Training and batch processing spend paid account credits.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/roboflow-agent-pluginDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The MCP server works with a Roboflow account and manages projects, datasets and training.
- Training and batch processing spend paid account credits.
This is third-party code. Review the repository files before installing.
What it does
The plugin bundles a set of skills with the official Roboflow MCP server configuration. The skills cover dataset management, model training and evaluation, inference (serverless, dedicated, self-hosted, batch), Workflows, the Universe catalogue, plans and an API reference. Separate skills draw annotations over images and video with supervision, upload custom weights and connect cloud storage. The `workflow-from-eval` skill builds a Workflow in an eval-first way: from a business goal and a set of frames the agent assembles test cases, compares candidates and tunes the winner. Packages are built for Claude Code, Codex, Microsoft Copilot Cowork and ChatGPT, and Cursor, OpenCode and other agents install the skills with `npx skills`.
Who it is for. Computer vision engineers and teams that train and deploy models on Roboflow and want an agent to drive the platform.
Good fit when
- Images must be uploaded to a Roboflow project, annotated and sent to training.
- A model is trained and an agent compares deployment options: serverless, dedicated, self-hosted or batch.
- A saved Workflow needs a test on video through the interactive preview.
- Detection needs ready-made models and datasets from the Universe catalogue.
Not a fit when
- The work does not run on Roboflow: the skills and MCP are tied to its platform and API.
- You only need to draw annotations and there is no reason to connect a platform account.
- No Roboflow account is available and neither OAuth sign-in nor an API key can be obtained.
Example request
Take my warehouse photos, suggest an annotation scheme and show whether to deploy the detection model in the cloud or on my own server.Limitations
The plugin works with a Roboflow account. The bundled MCP server uses OAuth sign-in: on first use the client asks you to authorize on Roboflow. Clients without OAuth need a private API key in the `x-api-key` header: setting `ROBOFLOW_API_KEY` alone is not enough. The ChatGPT and Copilot Cowork packages are built with scripts from the `chatgpt/` and `cowork/` folders, and some skills are left out of the ChatGPT variant. Some platform features depend on the plan, and training, inference and batch processing use account credits.
How to disable. Remove the `roboflow` plugin through the agent plugin manager or disable skills installed with `npx skills`. Revoke the MCP client's access in Roboflow account settings.
FAQ
How do I install the Roboflow plugin in Claude Code?
Run `claude plugin marketplace add roboflow/computer-vision-skills`, then `claude plugin install roboflow`. Add `--scope local` to install for the current project only.
Is an API key required?
No, if the client can complete OAuth: Roboflow sign-in is requested on first MCP use. Clients without OAuth need a key sent in the `x-api-key` header to `https://mcp.roboflow.com/mcp`.
How do I install a single skill?
`npx skills add roboflow/computer-vision-skills --skill roboflow-inference` installs one skill. Without `-g` it goes to `./.claude/skills/`, with `-g` to `~/.claude/skills/`.
Does the plugin work in Codex?
Yes. The repository has a Codex manifest `.codex-plugin/plugin.json` and a marketplace `.agents/plugins/marketplace.json`, both pointing to the same `skills/` folder.
What does the `draw-vision` skill do?
It guides drawing annotations, labels and predictions over images and video with supervision-js in JavaScript and TypeScript or Supervision in Python.
Discussion
No questions or reviews yet.
Related
PluginsTen MLflow skills teach a coding agent to add tracing, diagnose failures and evaluate GenAI apps
The intelli-flows plugin teaches Claude Code and Codex to write, run and explain multi-model Intelli flows.
AI Builder Club Skills
AI Builder Club — Skills
A Claude Code plugin marketplace to make a repo agent-ready and run compounding agent loops
A visual workflow builder for agents: draw a flow on a canvas and export it as skills and commands for Claude Code, Codex, Cursor and others
Declarative DAG workflows for coding agents: isolated phases, graph verification, resume and replay of runs.
A tool that builds tested skills from a real workflow and governs their lifecycle through your own marketplace
README in short
The README calls the repository the source of truth for the Roboflow plugin: skills live in `skills/`, the Claude Code and Codex manifests point to them, and the ChatGPT and Cowork packages are built from the same files. Other agents are covered by the `npx skills` route.