AmphiLoop

AmphiLoop

English | 中文

AmphiLoop, short for Amphibious Loop, is a new methodology, technology stack, and toolchain for building AI agents. It enables tasks to be described and orchestrated using natural language, with an “Explore → Code → Verify” loop guiding code generation and build. The resulting artifacts are capable of automatically switching between workflow mode and agent mode at runtime.

AmphiLoop packages domain knowledge and execution methodology into three layers:

Layer	Role	Description
Skills	Domain knowledge	"What it is, how to use it" — reference docs loaded on-demand
Agents	Execution methodology	"How to do it well" — specialized execution experts
Commands	Orchestration	Multi-step workflows that coordinate agents and skills

Together, they enable an end-to-end pipeline: explore a website via CLI -> generate a dual-mode agent project -> verify execution — all within an agent.

Further Reading

Long-form pieces on the motivation and design philosophy behind AmphiLoop:

Beyond Autonomous: Why I'm Building an Amphibious Agent

Installation

# Step 1: Register the marketplace (one-time)
claude plugin marketplace add bitsky-tech/AmphiLoop

# Step 2: Install the plugin
claude plugin install AmphiLoop

Or install from a local checkout (point marketplace add at the local directory — it's read as a marketplace because the repo ships .claude-plugin/marketplace.json):

git clone https://github.com/bitsky-tech/AmphiLoop.git
claude plugin marketplace add /path/to/AmphiLoop
claude plugin install AmphiLoop

After installation, skills, agents, and commands (e.g. /build) are automatically available in Claude Code.

Usage

Commands

Commands are user-invocable workflows. Invoke them with the / prefix:

/AmphiLoop:build

Unified pipeline. Describe any task, list the domain references the agents should read (SKILLs, CLI help, SDK docs, style guides), and ask to generate a runnable project:

/AmphiLoop:build

I want to aggregate all `orders_*.csv` files under ~/data/inputs into a single
summary.csv — one row per customer with totals.

Domain flag (optional) — append --<domain> to inject pre-distilled domain context from domain-context/<domain>/. Currently supported: --browser.

/AmphiLoop:build --browser

Go to https://example.com, search for "product", and extract the first 5 results.
I want a project that can run this reliably.

Without a flag, /build auto-detects the domain from TASK.md (and falls back to a generic flow if none matches). Users can always supply additional domain references in TASK.md.

What happens under the hood:

1. Initialize Task — Writes a TASK.md template where you fill in goal, expected output, and Domain References; auto-detects the domain if no flag was given
2. Configure Pipeline — Project mode (Workflow vs Amphiflow), LLM config if needed, plus any domain-specific configuration (e.g. browser environment mode when --browser is active)
3. Setup Environment — Checks uv, runs uv init
4. Explore — Delegates to amphibious-explore agent, which reads your domain references and probes the environment
5. Generate — Delegates to amphibious-code agent to produce a complete project with all source files
6. Verify — Delegates to amphibious-verify agent to inject debug instrumentation, run the project, and validate results

Agents

Agents are execution specialists delegated by commands. They are not called directly by users but are orchestrated internally:

Agent	What It Does
amphibious-explore	Systematically explores a target environment via a domain-supplied toolset, produces an executable plan with stability-annotated operations and supporting snapshots
amphibious-code	Generates a complete bridgic-amphibious project from a task description and exploration report
amphibious-verify	Injects debug instrumentation, runs the project with monitoring, validates results, and cleans up

Skills

Skills are domain knowledge references that agents and Claude load automatically when relevant. You don't invoke them directly — they activate based on conversation context.

Skill	Activates When
bridgic-browser	Using browser automation via CLI (bridgic-browser ...) or Python SDK (from bridgic.browser)
bridgic-amphibious	Building dual-mode agents with AmphibiousAutoma, CognitiveWorker, on_agent/on_workflow
bridgic-llms	Initializing LLM providers (OpenAILlm, OpenAILikeLlm, VllmServerLlm)

Architecture