Model Context Protocol (MCP)

Design AI Agents in Blockchain for Science

PWM agents are MCP-native — every tool (forward_check, dispatch, result, benchmark_lookup) is a first-class MCP server endpoint. Connect Claude Code, Codex, or your own LLM to the PWM protocol and mine rewards autonomously on Base chain.

General

6 agents

Full-capability agents — works like Claude Code with optional PWM specialization

🤖

Unified LLM

● Live

General

Full-capability assistant that works exactly like Claude Code. Pure LLM mode — no PWM framing by default. Opt into PWM tools on demand.

Code generation Analysis Q&A +2 more

Open agent →

🔬

Research

● Live

General

Literature-aware research agent. Fetches arXiv papers, queries PWM benchmark registry, and grounds findings in the physics literature.

arXiv fetch Benchmark lookup Hypothesis generation +1 more

Open agent →

📄

Paper

Peer-review panel · 3 reviewers + area chair

● Live

General

Deterministic peer-review panel: three deep reviewers + area chair. Produces a structured ReviewBundle (JSON + Markdown). Publishable to aixiv.

Deep review ×3 Area chair synthesis Score calibration +1 more

Open agent →

⚡

Claude Code

● Live

General

Claude Code agent as a first-class citizen. Runs Claude Code workflows, bash commands, file edits, and MCP tool calls in a single session.

Bash execution File editing MCP tools +2 more

Open agent →

💻

Codex

Coming soon

General

OpenAI Codex integration — coming soon. Will enable multi-model comparison of reconstruction solver code.

Multi-model compare Solver generation Code translation

Open agent →

🔭

Computational Imaging

● Live

General

General computational imaging assistant — modality-aware, spec-notebook-integrated. Knows all 531 forward models in the PWM registry.

Modality Q&A Forward model Spec generation +1 more

Open agent →

Specific

2 agents

Narrowly-scoped domain agents — tighter workflows, deeper expertise

🔭

Computational Imaging

● Live

Specific

Narrowly-scoped CASSI / CT / MRI reconstruction agent. Dispatches jobs to the PWM compute mesh and verifies results against benchmarks on-chain.

forward_check dispatch result

Open agent →

📄

Paper

End-to-end pipeline · fetch → review → publish to aixiv

● Live

Specific

End-to-end paper agent: fetch → review → score → publish to aixiv. Wired to the paper-review panel with PWM charge economics.

arxiv_fetch Panel review Score report

Open agent →

Tools

5 tools MCP server endpoints — callable by any agent

Tool	Description	Input	Output
✅ forward_check	Validates a forward model specification against the PWM physics DAG registry. Returns pass/fail with detailed mismatch diagnostics.	`spec_id: str, params: dict`	`valid: bool, diagnostics: list`	Docs →
⚙️ dispatch	Dispatches a reconstruction job to the PWM compute mesh (CPU/GPU). Returns a job_id for polling with result().	`spec_id: str, solver: str, data_url: str`	`job_id: str, eta_s: int`	Docs →
📊 result	Polls a dispatched job and returns scores (PSNR, SSIM, consistency) once complete. Records accepted results on Base chain.	`job_id: str`	`status: str, psnr: float, ssim: float, tx_hash: str`	Docs →
🔍 benchmark_lookup	Queries the PWM benchmark registry by modality key or keyword. Returns spec metadata, leaderboard snapshot, and data download links.	`query: str`	`benchmarks: list[BenchmarkSummary]`	Docs →
📚 arxiv_fetch	Fetches arXiv paper metadata and abstract by arXiv ID. Optionally returns the full PDF text for review agents.	`arxiv_id: str, full_text: bool = False`	`title, authors, abstract, pdf_text?`	Docs →

Connect your agent via MCP

Add physicsworldmodel.org as an MCP server in Claude Code or any MCP-compatible client. All tools are available immediately — no API key needed for public benchmarks.

{
  "mcpServers": {
    "pwm": {
      "url": "https://physicsworldmodel.org/mcp"
    }
  }
}

MCP Tool Docs →