How It Works

Four steps from plug-in to verified code.

1Connect

Add The 35 as an MCP server in your Claude Code config. Your agent gains compliance tools alongside its usual capabilities — compliance_check, known_patterns, cross_reference, deep_compliance_check, sdk_audit.

{
  "mcpServers": {
    "the-35": {
      "type": "http",
      "url": "http://localhost:4000/mcp"
    }
  }
}

Fund a wallet. Point it at the gateway. Your agent can now pay for knowledge as part of its normal workflow.

2Query

As you code, your agent checks implementations against specs, looks up known failure patterns, and cross-references other SDKs. Each query is a BSV micro-payment, settled on-chain in the background.

Agents use The 35 for things like:

• "Is this function compliant with BRC-42 §3.2?" — cascades to library + findings + reference
• "What failure patterns exist for sighash implementations?" — hits findings service
• "How do TS, Go, Ruby handle this exact operation?" — hits reference service
• "What coverage does the codebase currently have?" — hits coverage service

The agent barely notices it's using an external service. It just has better tools and produces more compliant code.

3Review

Run a full compliance review on demand. The triumvirate teams audit each phase of the codebase. The ten-specialist council votes on every finding. An independent auditor tracks coverage.

Output:

• Compliance review document — findings with severity, quorum votes, recommended actions, and fix patches
• Coverage report — traffic-light (GREEN/YELLOW/RED) percentages across the codebase, per-file status
• Provenance metadata — for every finding, which sources were consulted and how

The review is itself a cascade of thousands of paid queries flowing through the service mesh. Each triumvirate's work generates real transactions against real data. You pay for the audit, the audit pays its constituents, the constituents pay their data sources.

4Verify

Get a compliance coverage report: what percentage of your codebase has been semantically verified against spec. GREEN, YELLOW, RED — at a glance.

The metric is honest about what it knows. Unlike test coverage, it distinguishes between "this code was executed" and "this code was semantically validated." Two different questions, two different answers.

What you get

Three things no other tool provides:

1. Real-time compliance intelligence as you code — through MCP tools the agent uses naturally
2. Full audit capability on demand — for PRs, releases, or just before you ship
3. Measurable compliance coverage — a semantic metric you can track over time like test coverage, but for correctness rather than execution

All paid in BSV. All cascading through the service mesh. All producing verifiable on-chain records of what knowledge you consumed and when.