Convex Agent Plugins
The official Convex plugin for Cursor and Claude Code: rules, commands, MCP access and hooks for TypeScript backends.
- 117 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The deploy key gives the MCP server access to data and functions of a live Convex deployment.
- The `/add` command downloads scripts and runs them on your machine.
Install
- Works in
Manual install
cd ~/.cursor/plugins && git clone https://github.com/convex/convex-agent-plugins convexManual install from the README with the repository URL filled in for the placeholder. Restart Cursor after cloning, or install through the Cursor Marketplace.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The deploy key gives the MCP server access to data and functions of a live Convex deployment.
- The `/add` command downloads scripts and runs them on your machine.
This is third-party code. Review the repository files before installing.
What it does
The Convex plugin teaches an agent to build backends on Convex, a reactive platform with a database and server functions in TypeScript. Rules apply on their own while the agent works in the `convex/` directory: indexes instead of `.filter()`, argument and permission checks, flat schemas. Separate commands run common tasks: `/add`, `/migrate`, `/seed`, `/test`, `/deploy-guard`, `/ship`, `/optimize` and others. Two agents, `convex-advisor` and `convex-reviewer`, advise on architecture and review code. The manifest wires up the MCP server `npx -y convex@latest mcp start`, which gives access to the schema, configuration and functions of a deployment. Cursor hooks block a `git commit` with `Date.now()` or `.filter()` in queries and, after an agent turn, run `npx tsc --noEmit` and hand any errors back to the agent.
Who it is for. Convex developers who write backends with Cursor or Claude Code and want the code kept to recommended practices.
Good fit when
- A Convex project starts from scratch and the agent has to prepare a schema, permissions and the first functions.
- A schema changes on live data and the migration must run without downtime.
- You want to catch queries that break reactivity automatically before a commit.
- Backend code is reviewed by a reviewer who knows common Convex mistakes.
Not a fit when
- The backend is not built on Convex.
- The deploy key cannot be handed to an agent.
- You need a hard block on ending a turn when the build fails: in Cursor the `stop` hook only returns errors to the agent.
Example request
Prepare a Convex schema for a task manager with users, teams and tasks and add a task update mutation with an ownership check.Limitations
The plugin is built for Cursor: hooks follow the Cursor format, and the README names Claude Code as a second supported agent without separate instructions for it. MCP needs `CONVEX_DEPLOYMENT` and `CONVEX_DEPLOY_KEY`. The `stop` hook repeats a turn at most twice and cannot forbid finishing. The `/add` command downloads a capability catalog and setup scripts from the Convex service and runs them, and each download is logged as an anonymous event. Without `/add` the plugin sends nothing.
How to disable. Delete the plugin directory from `~/.cursor/plugins` and restart Cursor. Remove `CONVEX_DEPLOY_KEY` from the environment.
FAQ
How do I install the Convex plugin in Cursor?
Through the Cursor Marketplace, or manually: clone the repository into `~/.cursor/plugins/convex` and restart Cursor.
Which variables does the MCP server need?
`CONVEX_DEPLOYMENT` with the deployment name and `CONVEX_DEPLOY_KEY`. The key comes from the Convex Dashboard.
What do the plugin hooks do?
Before `git commit` a script looks for `Date.now()` in queries and `.filter()` after `db.query(...)` and can reject the commit. After an agent turn, if a `convex/` directory exists, `npx tsc --noEmit` runs and any errors go back to the agent as the next message.
Which commands does the plugin have?
Among them are `/add`, `/migrate`, `/seed`, `/test`, `/deploy-guard`, `/ship`, `/optimize`, `/monitor`, `/crons` and `/env`.
What does the `/add` command do?
It downloads a capability catalog and setup scripts from the Convex service and runs them. Each download is logged as an anonymous event; without `/add` the plugin sends nothing.
Discussion
No questions or reviews yet.
Related
PluginsConvex skills for agents: project setup, authentication, components, migrations and performance investigation
builder-skills: agent skills for Convex apps
builder-skills
17 skills that teach an agent to build Convex apps by the docs and run a project through a PRD, task.md and a changelog
The official Tencent CloudBase kit for coding agents: a plugin installs the stack, skills set the rules, MCP runs the backend
An MCP server and plugin for agents that generate web APIs from OpenAPI, AsyncAPI, GraphQL, gRPC and SQL specs on a Rust core.
A PostgreSQL backend your agent builds over MCP, where the platform verifies every change and can roll it back.
Plinth
plinth
Skills and agents for Java development: build, tests, profiling and architecture. One-command installation for Claude Code, Cursor, Codex and Copilot
README in short
The README splits the plugin into rules, skills, agents, MCP and hooks. A separate section explains how the Cursor hook differs from the blocking Claude Code hook.