cc-fleet
A CLI that plugs third-party models into Claude Code orchestration: workflows, agent teams and subagents on DeepSeek, GLM, Kimi, Qwen or a Codex subscription
High risk
We rate an entry high when the tool writes to external systems, handles money, production databases or secrets, or runs arbitrary commands. The CLI installs it only with your consent.
Why this level
- Installs and updates a binary from the network
- Spawns agent processes on third-party providers
- Handles provider keys and subscriptions
Install
Manual install
curl -fsSL https://raw.githubusercontent.com/ethanhq/cc-fleet/main/install.sh | shInstalls the binary with a ccf alias and the Claude Code plugin. Claude Code itself must be installed first.
This is third-party code. Review the repository files before installing.
What it does
Claude Code orchestration, that is dynamic workflows, agent teams and subagents, runs only on Anthropic models by default. cc-fleet lets any model with an Anthropic- or OpenAI-compatible API join, down to a Codex subscription. Every third-party worker is a real claude process with its model backend swapped, so Claude Code drives it like a native agent. Provider keys never enter env, argv or shell history, and the main session's own auth is untouched. It installs with one command and ships an interactive TUI plus a plugin with skills and a session hook.
Who it is for. For developers who want to run Claude Code multi-agent flows on third-party or cheaper models.
Good fit when
- You need to run Claude Code workflows and subagents on a non-Anthropic model
- You want to use a Codex subscription as the backend for agents
- You want an interactive claude on a chosen provider without an Anthropic subscription
Not a fit when
- The stock Claude Code orchestration on Anthropic models is enough
- You cannot install a third-party binary and update it from the network
Example request
Run subagents on DeepSeek to review these three modules in parallelLimitations
The tool drives the official claude CLI, so that has to be installed separately. The install script downloads and verifies the binary, and there is also an npm package. It needs third-party provider keys or subscriptions, whose limits and availability sit with the provider. Availability of some providers and of Claude Code itself from Russia without a VPN is not guaranteed. Agent teams stay Claude Code only; from Codex you get subagents and workflows.
How to disable. Run ccf uninstall --all to remove the binary and the plugin.
Security check
- Installs and updates a binary from the network
- Spawns agent processes on third-party providers
- Handles provider keys and subscriptions
README in short
The README explains that Claude Code orchestration runs only on Anthropic models by default, and cc-fleet lifts that limit. Any model with an Anthropic- or OpenAI-compatible API joins as a workflow leaf, a long-lived teammate or a one-shot subagent, and every worker is a real claude process with a swapped backend. Install is two steps: a script sets up the ccf binary and plugin, then you register a provider. Control comes from Claude Code with /workflow, /team, /subagent, and directly via ccf; Codex is supported both as a backend and as a host.
FAQ
Do I need an Anthropic subscription?
Not to run on a third-party provider: ccf run on a chosen model gives an interactive claude billed to the provider key. The claude CLI itself still has to be installed.
Do provider keys leak?
The author states that keys never enter env, argv or shell history, and the main session's auth is untouched.
Related
Open-source personal AI assistant on your own machine: answers in Telegram, Slack, Discord and WhatsApp, extended with skills and plugins
A self-improving agent from Nous Research with a TUI, messaging gateway, cron jobs and skills it writes itself
An open source coding agent for the terminal and desktop with build and plan modes
Open prompt library with a Claude Code plugin, MCP server and CLI: search, fetch and improve prompts and skills from an agent