# OPSIN

> OPSIN is a free MCP server for ai models and llm tooling. Signed, preregistered experiments on which prompts and tools help AI agents. It runs as a hosted remote server, so there is nothing to install locally.

- **Kind:** MCP server
- **Pricing:** Free
- **Platforms:** API, Web
- **Website:** https://opsinlab.io
- **MCP Registry name:** io.opsinlab/opsin
- **Categories:** AI models and LLM tooling
- **Listing updated:** 2026-11-14

## About

OPSIN runs an automated lab that toggles individual parts of an AI agent, such as a calculator tool or a prompt instruction, and measures what changes against two controls and a placebo. Each result is a signed verdict marked as exciting, inhibiting, inert or unresolved, with a confidence interval and the number of paired runs. Tasks are drawn by a public randomness beacon and plans are preregistered.

The site reports 767 signed verdicts across 124 notebooks and 545 parts measured. In its first notebook, Claude Haiku 5.5 acting as a trading agent's maths module solved 432 DeFi problems over 14,112 runs, and a calculator tool raised exact answers far more than prompt tricks such as personas or step-by-step wording. Visitors can also try their own instruction on the lab's budget, three free trials a day.

The remote MCP endpoint gives agents access to this record. Preregistrations, per-run CSVs, verdict JSON and the code to rerun experiments are published.

## Features

- **Signed verdicts:** Every result is signed and labelled excites, inhibits, inert or unresolved.
- **Preregistered notebooks:** Plans are fixed in advance and tasks drawn by a public beacon.
- **Controls and placebo:** Effects are measured against two controls and a placebo.
- **Open data:** Per-run CSVs, verdict JSON and rerun code are published.
- **Try your own line:** Test an instruction against Claude Haiku 5.5 on the lab's budget.
- **On-chain identity:** OPSIN lists an ERC-8004 validator identity on Base.

## Good for

- Check whether a prompt trick helps before adopting it
- Compare a tool against a prompt instruction for maths accuracy
- Look up evidence on a specific agent skill
- Rerun a published experiment
- Back an agent-hiring decision with measured effects

## Install

### Claude Code (Terminal)

```sh
claude mcp add --transport http opsin https://opsinlab.io/mcp
```

### Claude Desktop (Settings → Connectors → Add custom connector)

```sh
https://opsinlab.io/mcp
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "opsin": {
      "url": "https://opsinlab.io/mcp"
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "opsin": {
      "type": "http",
      "url": "https://opsinlab.io/mcp"
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "opsin": {
      "serverUrl": "https://opsinlab.io/mcp"
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "opsin": {
      "type": "http",
      "url": "https://opsinlab.io/mcp"
    }
  }
}
```

## FAQ

### What does OPSIN measure?

Which parts of an agent change exact-answer scores, using paired runs with controls.

### Can I test my own instruction?

Yes, three free trials a day, each running three rounds of eight problem pairs.

### Are small effects detected?

Not always; the site says 24 pairs only settle big effects and small ones stay unresolved.

### Is the data public?

Preregistrations, every run as CSV, verdicts as JSON and the code are published.

### Is OPSIN free?

Yes. OPSIN is free to use.

### How do I add OPSIN to Claude Code?

Run this command in your terminal: claude mcp add --transport http opsin https://opsinlab.io/mcp

### What are the best alternatives to OPSIN?

The closest alternatives to OPSIN are Claude Code Ultimate Guide, signafAIde, Toolkien, Lugh and HypePaper.

## Alternatives

- [Claude Code Ultimate Guide](https://ai-powered.com/mcp/claude-code-ultimate-guide): Claude Code Ultimate Guide is an open-source MCP server for ai models and llm tooling.
- [signafAIde](https://ai-powered.com/mcp/signafaide): signafAIde is a freemium MCP server for security.
- [Toolkien](https://ai-powered.com/mcp/toolkien): Toolkien is an open-source MCP server for ai models and llm tooling.
- [Lugh](https://ai-powered.com/mcp/lugh): Lugh is an open-source MCP server for ai models and llm tooling.
- [HypePaper](https://ai-powered.com/mcp/hypepaper): HypePaper is a freemium MCP server for ai models and llm tooling.
- [Agent Data Tools: Compact (find, describe, run)](https://ai-powered.com/mcp/agent-data-tools-compact-find-describe-run): Agent Data Tools: Compact (find, describe, run) is a freemium MCP server for developer tools.
- [Angel Agent Guard: pre-purchase check + prompt-injection scan](https://ai-powered.com/mcp/angel-agent-guard-pre-purchase-check-prompt-injection-scan): Angel Agent Guard: pre-purchase check + prompt-injection scan is a freemium MCP server for ai models and llm tooling.
- [Mizan Calibration](https://ai-powered.com/mcp/mizan-calibration): Mizan Calibration is a free MCP server for ai models and llm tooling.

---
Source: https://ai-powered.com/mcp/opsin
