CLI & AI integrations overview
The Convoy CLI is the bridge between the web app and your machine — live threads, MCP config, skills, and task scripting.
The Convoy CLI (convoy) is the local half of the product. It has two
roles:
- a long-running agent bridge (
convoy connect) that executes AI threads sent from the web app on your machine - a one-shot tool for tasks, projects, MCP config, and skill installation from the terminal
The three verbs
Everything in this section hangs off three commands:
| Command | What it does | When you run it |
|---|---|---|
convoy setup | Saves a profile — deployment URL + API key + project + local paths | Once per project/machine |
convoy connect | Runs the long-lived bridge that claims and executes AI threads | Whenever threads should execute on this machine |
convoy init | Writes MCP config and installs skills for AI clients (Claude Code, Codex, OpenCode) | Once per repo you want wired into editors/agents |
connect is only needed for live thread execution — MCP and skills work
without it.
Reading order
The pages in this section build on each other:
API keys— the credential everything below usesInstall the CLISetup and connect— profile + bridgeProfiles, paths, and status— multiple projects, worktrees, resolution rulesModel provider keys— optionally pin a profile to Bedrock or MantleSet up MCP— expose Convoy tasks to MCP-aware clientsSkills— teach AI clients the Convoy task workflow
Then, as needed:
One-shot commands— scripting tasks from the terminal (full surface in theCLI command reference)Security model— what keys grant and what a connected agent can do; read this before connecting a shared projectDev container— isolated environment for running AI CLIsCLI troubleshooting