Score breakdown
Popularity is tracked separately. Support, ads, sponsorships, and tips never affect these signals.
Why it matters
Useful for engineering teams running Claude Code who want to mix-and-match LLMs in a single workflow without giving up the Claude Code UX: cc-fleet is the Apache-2.0 CLI + plugin that lets any Anthropic- or OpenAI-compatible backend (DeepSeek, GLM, Kimi, Qwen, OpenRouter, your Codex subscription) join Claude Code's Dynamic Workflows / Agent Teams / Subagents as a workflow leaf / long-lived teammat
Who should use it
Who should skip it
Skip ethanhq/cc-fleet unless the captured evidence suggests it solves a problem you are actively working on.
About this signal
ethanhq/cc-fleet is tracked by RepoRadar as a plug any 3rd-party llm into clau in the Apache-2.0 npm-published CLI + Claude Code plugi section. It was first seen on 2026-06-25 and last updated on 2026-06-25. The current verdict is 'try now' with a Silver tier and easy setup difficulty. The standout signals for ethanhq/cc-fleet are workflow potential (9.0) and setup ease (8.8), while momentum (7.0) trails — that balance shapes where it fits best. This page summarizes the evidence RepoRadar has captured from captured source metadata. The score, tier, risk label, and verdict on this page are never influenced by sponsorship, ads, or tips — they reflect only the usefulness, popularity, novelty, momentum, maturity, and evidence signals described in the RepoRadar methodology.
How this item is evaluated
RepoRadar assigned ethanhq/cc-fleet a composite score of 7.9 out of 10, placing it in the Silver tier. This score combines weighted sub-signals: usefulness (35%), novelty (18%), momentum (14%), maturity (10%), open-source/build quality (7%), evidence quality (6%), workflow potential (6%), and setup ease (4%). Popularity is tracked separately at 162.0 and never affects the composite score or tier. The risk label of 'low' reflects inherent user-impacting hazards, not generic novelty. Items with no risk flag may still require normal code review before production use.
Risk explanation
**Provider keys live in a protected store, but the store itself is a security boundary.** cc-fleet's zero-leak design means provider keys never enter env / argv / shell history, but the protected key store is itself a security boundary that needs the same protection as the agent's own credentials. Adopters should review the store's on-disk encryption + access controls (and the install path's filesystem permissions) before registering production provider keys. The keys are not exposed to the agent harness, so a compromised Claude Code session cannot exfiltrate them, but a compromised host with read access to the store can; **Frontier-model swap changes cost / quality / latency; benchmark before betting production routing on it.** cc-fleet makes it easy to route cheap tasks to a cheap provider and expensive tasks to a frontier provider, but the orchestrator's task-to-provider routing is only as good as the team's understanding of what each model is good at. A 7B model routed to a deep-reasoning task will fail; a frontier model routed to a lint task will burn money. Adopters should benchmark every registered provider against representative tasks (`ccf --benchmark`) and tune the routing deliberately — `--cost` shows the per-provider price table but does not pick the right provider per task; **Chinese-provider worker composition (GLM / Qwen / Kimi) is first-class; verify model identity and license terms for the team's jurisdiction.** The README's 简体中文 version explicitly mentions GLM / Qwen / Kimi as first-class workers, and the cross-border-data implications of routing Claude Code workflows through Chinese-provider workers depend on the team's jurisdiction and data-residency requirements. Adopters in regulated environments (EU GDPR, US HIPAA, China PIPL, etc.) should verify that the chosen provider's data-residency + model-identity + license terms match the team's compliance posture before registering the provider as a worker.