Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Multi-model orchestration layer for Claude Code — the frontier model plans, cheaper models execute, verification guards quality. One-prompt install.
| Date | Stars |
|---|---|
| 2026-07-31 | 548 |
| 2026-08-05 | 581 |
| 2026-08-06 | 581 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
60.0
growth rate 0.00%/day
# pilotfish 🐟 > Pilot fish swim alongside the ocean's largest predators — small, fast, and doing the routine work so the big one doesn't have to. **pilotfish** is a multi-model orchestration layer for [Claude Code](https://code.claude.com): the `opus` family plans and decides in your main session, Sonnet and Haiku execute the volume work through global subagents, and fresh Opus contexts challenge Plans and completed outcomes. Quality is protected by independent verification, not by using the biggest model everywhere. Everything installs globally — one setup, every project — and the whole stack degrades gracefully when the primary model becomes unavailable. > **Want OpenAI GPT-5.6 inside Claude Code without changing native Claude state?** [remora](https://github.com/Nanako0129/remora-cc) packages pilotfish's role-based orchestration pattern into a session-scoped launcher for an existing Anthropic-compatible gateway. Use pilotfish to study or customize the global policy; use remora for an approval-gated, verifiable install whose model and gateway overrides disappear with the child process. > **Want the same orchestration on Grok Build?** [pilotfish-grok](https://github.com/Nanako0129/pilotfish-grok) ports the role lifecycle and capability boundaries to `~/.grok/` (agents, roles, and a model-free policy). Use this repo for Claude Code; use pilotfish-grok when the host is Grok Build — separate install surface, does not write `~/.claude/`. **Where this came from:** my weekly quota reset one morning, and the first thing I did with a fresh Fable 5 allowance was ask it to figure out why the previous week's had evaporated. This repo is the setup that research produced, and it's what I now run daily on every project — three config files, no runtime code. The research notes (with sources) are in [docs/](./docs/). [繁體中文說明](./README.zh-TW.md) ## Contents - [Why](#why) - [How it works](#how-it-works) - [Install](#install) - [Trust & security](#trust--security) - [What gets installed](#what-gets-installed) - [Updating](#updating) - [The fallback story](#the-fallback-story) - [Tuning & FAQ](#tuning--faq) - [Research & design](#research--design) - [Contributing](#contributing) - [Uninstall](#uninstall) - [Support pilotfish](#support-pilotfish) - [License](#license) ## Why On 2026-07-24, Anthropic released [Opus 5](https://www.anthropic.com/news/claude-opus-5), describing it as close to Fable 5 intelligence at half the API price. Opus 5 leads many of Anthropic's published evaluations, but not every one. pilotfish therefore defaults **new installs** to the `opus` family alias and keeps Fable 5 as an explicit `/model fable` opt-in. This is a cost-aware default, not a claim that Opus 5 is universally better; the decision and rollback criteria are tracked in [#23](https://github.com/Nanako0129/pilotfish/issues/23). The original July research still explains the architecture: frontier-model sessions are expensive, while most coding-session tokens are searching, mechanical edits, test runs, and doc updates rather than judgment. Those high-volume paths can use Sonnet or Haiku while acceptance-boundary reviews use a fresh Opus context. Every piece of this now carries Anthropic backing. The [Fable 5 prompting guide](https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/prompting-claude-fable-5) recommends frequent subagent delegation and notes that **independent fresh-context verifier subagents outperform self-critique**. And as of 2026-07-08, the cheap-executor split is officially benchmarked: Anthropic's own tests put a **Fable 5 orchestrator with Sonnet 5 workers at 96% of all-Fable performance for 46% of the cost** (BrowseComp: 86.8% vs 90.8% accuracy, $18.53 vs $40.56 per problem), with the inverse advisor pattern (Sonnet executor consulting Fable) at ~92% for ~63% on SWE-bench Pro — the orchestrator split pilotfish uses won on both axes ([multi-agent docs](https://platform.claude.com/docs/en/managed-agents/multi-agen
Excerpt of 36,622 characters
Read on GitHub
Nyanako · Taiwan
145
2
2
1
1
1
Po-Han Shih
1
Would you bet a product on this? Bounded 0–100 and slow moving.
matched fp:1b95e7e51f5f93ae, topic:ai-agents, topic:multi-agent
matched fp:1b95e7e51f5f93ae, topic:orchestration