Conduit MCP Server

Local setup required. This server has to be cloned and prepared on your machine before you register it in Claude Code.
1

Set the server up locally

Run this once to clone and prepare the server before adding it to Claude Code.

Run in terminal
pip install conduit-browser
2

Register it in Claude Code

After the local setup is done, run this command to point Claude Code at the built server.

Run in terminal
claude mcp add conduit -- node "<FULL_PATH_TO_CONDUIT>/dist/index.js"

Replace <FULL_PATH_TO_CONDUIT>/dist/index.js with the actual folder you prepared in step 1.

README.md

The only headless browser with a cryptographic audit layer.

Conduit

The only headless browser with a cryptographic audit layer.

Every action Conduit takes — every click, every navigation, every JavaScript execution — is written to a tamper-evident SHA-256 hash chain, signed with an Ed25519 identity key, and verifiable by anyone with zero dependencies. No other headless browser does this.


Install

pip install conduit-browser

Or from source:

git clone https://github.com/bkauto3/Conduit.git
cd Conduit
pip install -r requirements.txt

Quick Start — Audited Session in 60 Seconds

import asyncio
from tools.conduit_bridge import ConduitBridge

async def main():
    bridge = ConduitBridge()

    # Navigate to a page
    result = await bridge.execute({"action": "navigate", "url": "https://example.com"})
    print(result["title"])

    # Extract main content (strips nav/ads/footers)
    content = await bridge.execute({"action": "extract_main", "fmt": "md"})
    print(content["text"])

    # Export cryptographic proof of the entire session
    proof = await bridge.execute({"action": "export_proof"})
    print(f"Proof bundle: {proof['path']}")
    print(f"Verify: cd session_proof && python verify.py")

asyncio.run(main())

Use Cases

Compliance automation — Prove a specific form was filled with specific values at a specific time. Export a proof bundle. The chain hash is your receipt.

Security research — Document what JS a page injected, what network requests it made, what the DOM looked like at each step — all signed and chained.

AI agent browser control — Designed as the browser engine for autonomous agents. Budget enforcement prevents runaway costs. The audit trail lets you replay and inspect exactly what the agent did.

Web monitoringfingerprint + check_changed gives you signed change detection with cryptographic proof of when a page mutated.

Site mapping and bulk extraction — BFS crawl with robots.txt compliance, adaptive rate limiting, and per-page audit events.


Built for Agent Economies

Conduit's audit trail is not just for compliance — it is the trust layer that enables agents to transact with each other. When Agent A hires Agent B to do web research, the proof bundle is how Agent A knows the work was actually done.

This is the model behind SwarmSync.ai, an agent marketplace where 420+ agents negotiate, execute, and get paid — with Conduit providing the verifiable execution layer. Conduit is and will always be free and open-source. SwarmSync is where the work gets monetized.

You do not need SwarmSync to use Conduit. But if your agent does useful web work, SwarmSync is where other agents will find it and pay for it.


For Compliance & Legal Teams

Conduit proof bundles serve as chain-of-custody documentation for web-based evidence:

  • SOC 2 / SOX audits — Prove exactly what automated systems did during testing and monitoring (CC7.2 change monitoring, CC6.1 logical access)
  • GDPR verification — Document that a site deleted personal data or displayed required consent banners, with timestamped proof
  • Litigation support — Capture what a website displayed at a specific moment, with tamper-evident chaining that holds up to scrutiny
  • Insurance claims — Document property listings, damage reports, or policy terms with cryptographic proof of capture time
  • HIPAA audit trails — Prove exactly which automated processes accessed what data and when (164.312(b) audit controls)

Each proof bundle is self-verifiable with zero dependencies and can be archived alongside your compliance records. Think of it as a notarized logbook where tearing out or altering any page makes the tampering obvious.


For Security Researchers

Full JavaScript Source in the Audit Chain

When you execute JavaScript via eval, Conduit stores the entire source body in the hash chain — not just the result:

result = await bridge.execute({
    "action": "eval",
    "js": "Array.from(document.scripts).map(s => s.src)"
})

This means you can:

  • Prove exactly which code executed on a page
  • Detect if a page injected unexpected scripts
  • Document web-based exploits with cryptographic evidence
  • Build forensic session replays where every action is signed and chained

No other headless browser captures the JS source itself — they only log that JS ran and what it returned. Conduit logs what ran.


Why Conduit Instead of Playwright, Puppeteer, or Selenium?

| Fea

Tools (4)

navigateNavigates the browser to a specified URL.
extract_mainExtracts the main content of the current page, stripping navigation, ads, and footers.
export_proofExports a cryptographic proof bundle of the current session.
evalExecutes JavaScript on the current page and records the source in the audit chain.

Configuration

claude_desktop_config.json
{"mcpServers": {"conduit": {"command": "python", "args": ["-m", "conduit.server"]}}}

Try it

Navigate to https://example.com and extract the main content in markdown format.
Execute a script to list all script sources on the current page and ensure the action is recorded in the audit trail.
Export a cryptographic proof bundle for the current browsing session to verify the actions taken.
Monitor the page for changes and provide a signed proof of the mutation.

Frequently Asked Questions

What are the key features of Conduit?

Tamper-evident SHA-256 hash chain for all browser actions. Ed25519-signed proof bundles for cryptographic verification. Full JavaScript source capture during evaluation. Stealth-enabled browser engine via Patchright. Built-in site crawling and page monitoring.

What can I use Conduit for?

Compliance automation for SOC 2 or SOX audits requiring verifiable logs. Security research documenting injected scripts and network requests. Autonomous agent browser control with budget enforcement and audit trails. Litigation support for capturing website state at a specific moment. GDPR verification of consent banners and data deletion.

How do I install Conduit?

Install Conduit by running: pip install conduit-browser

What MCP clients work with Conduit?

Conduit works with any MCP-compatible client including Claude Desktop, Claude Code, Cursor, and other editors with MCP support.

Turn this server into reusable context

Keep Conduit docs, env vars, and workflow notes in Conare so your agent carries them across sessions.

Need the old visual installer? Open Conare IDE.
Open Conare