MCP (HTTP) https://narrowhighway.com/mcp/
MCP (SSE) https://narrowhighway.com/mcp/sse/
REST · OpenAPI https://narrowhighway.com/openapi.json
Claude Desktop Cursor ChatGPT / OpenAI REST / curl Python SDK Local install (stdio)

1Claude Desktop

The fastest path. Hosted MCP — no install. Add three lines to your config and 86 tools light up in any new conversation.

Step 1 · Open the config file

  • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
  • Windows: %APPDATA%\Claude\claude_desktop_config.json

If the file doesn't exist, create it. If it has other servers, just add concordance alongside them.

Step 2 · Paste this

{
  "mcpServers": {
    "concordance": {
      "type": "http",
      "url": "https://narrowhighway.com/mcp/"
    }
  }
}

Step 3 · Restart Claude Desktop

Open a new conversation. Look for the tool icon in the input box — you should see 86 tools available. Try: "Use concordance to verify that 2 H2 + O2 → 2 H2O is balanced."

No install No auth needed All 86 tools
Anthropic MCP docs ↗

2Cursor IDE

Cursor's Settings → MCP panel takes a URL directly. Tools become available with @concordance in any chat.

Step 1 · Open MCP settings

In Cursor: Cmd/Ctrl + Shift + P → "Open MCP Settings", or Settings → Features → MCP.

Step 2 · Add the server

Name: concordance
URL:  https://narrowhighway.com/mcp/

Step 3 · Use it

In any chat: @concordance verify_chemistry "C3H8 + 5 O2 -> 3 CO2 + 4 H2O"

Same endpoint as Claude Desktop Tab-complete on tool names

3ChatGPT & OpenAI agents

Auto-import the OpenAPI spec — every endpoint becomes a function call. Works with Custom GPTs, Assistants API, and any framework that consumes OpenAPI.

Custom GPT (zero code)

  1. Open ChatGPT → "Explore GPTs" → "Create"
  2. Configure → Actions → "Import from URL"
  3. Paste: https://narrowhighway.com/openapi.json
  4. Authentication: None (read endpoints) — or add an API key for write endpoints

The Custom GPT now has every Concordance endpoint as a tool — verify, capture, seal, polymathic synthesis, journal queries.

OpenAI Assistants API (Python)

import openai, json, urllib.request

# Pull the OpenAPI spec — 240+ endpoints, all 86 MCP tools mirrored
spec = json.loads(urllib.request.urlopen(
    "https://narrowhighway.com/openapi.json").read())

client = openai.OpenAI()
assistant = client.beta.assistants.create(
    name="Concordance-backed agent",
    instructions="Verify any irreversible computation against the Concordance engine before answering.",
    model="gpt-4o",
    tools=[{"type": "function", "function": f}
            for f in openapi_to_tools(spec)],
)

Helper openapi_to_tools in the examples directory.

Live API docs (Swagger) ↗

4REST / curl

86 HTTP endpoints. Read endpoints (verifier inventory, grid, journal, health) need no auth. Write endpoints (capture, seal, ledger writes) accept an optional API key for attribution.

Verify a chemistry claim

# POST /verify — all 69 verifiers reachable
curl -X POST https://narrowhighway.com/verify/chemistry \
  -H "Content-Type: application/json" \
  -d '{"equation": "H2 + O2 -> H2O"}'

Recompute a p-value from raw inputs

curl -X POST https://narrowhighway.com/verify/statistics_pvalue \
  -H "Content-Type: application/json" \
  -d '{
    "test": "paired_t",
    "n": 20,
    "mean_diff": 0.5,
    "sd_diff": 1.0,
    "tail": "two",
    "claimed_p": 0.0375
  }'

Polymathic synthesis (cross-domain)

curl -X POST https://narrowhighway.com/polymathic \
  -H "Content-Type: application/json" \
  -d '{"text": "Our church is approving a $2M building loan at 7% over 30 years..."}'

Read endpoints (no auth)

curl https://narrowhighway.com/health        # status + module counts
curl https://narrowhighway.com/identity      # engine identity statement
curl https://narrowhighway.com/llms.txt      # tool inventory for AI crawlers
curl https://narrowhighway.com/verifiers     # list all 69 verifiers
curl https://narrowhighway.com/journal/recent
curl https://narrowhighway.com/grid/scaffold # axis assignments per domain
curl https://narrowhighway.com/audit-chain   # sealed packets, newest first
Open Swagger UI ↗ Open ReDoc ↗

5Python SDK

For agents you're writing yourself. Three lines to a working verifier.

Direct REST (no dependency)

import requests

r = requests.post(
    "https://narrowhighway.com/verify/chemistry",
    json={"equation": "H2 + O2 -> H2O"},
)
print(r.json())
# {"checks":[{"status":"MISMATCH","detail":"balances as: 2 H2 + O2 -> 2 H2O", ...}]}

MCP client (for full tool access)

from mcp.client.streamable_http import streamablehttp_client
from mcp import ClientSession

async with streamablehttp_client("https://narrowhighway.com/mcp") as (r, w, _):
    async with ClientSession(r, w) as s:
        await s.initialize()
        tools = (await s.list_tools()).tools
        result = await s.call_tool("verify_chemistry",
                                      {"equation": "H2 + O2 -> H2O"})
        print(result)

LangChain

from langchain_mcp_adapters.client import MultiServerMCPClient
from langgraph.prebuilt import create_react_agent

mcp = MultiServerMCPClient({
    "concordance": {"transport": "streamable_http",
                     "url": "https://narrowhighway.com/mcp"}
})
agent = create_react_agent("openai:gpt-4o", await mcp.get_tools())

6Local install (stdio)

For air-gapped, offline, or sovereignty-first deployments. The same engine runs on your machine — under 1 GB, no outbound calls. Use the four-mode deployment plan.

Install

git clone https://github.com/matharrismma/Lighthouse
cd Lighthouse
pip install -e ".[mcp]"

Configure Claude Desktop for the local stdio server

{
  "mcpServers": {
    "concordance": {
      "command": "concordance-mcp"
    }
  }
}

Or the dev-tree version (no install):

{
  "mcpServers": {
    "concordance": {
      "command": "python",
      "args": ["-m", "concordance_engine.mcp_server"],
      "env": {"PYTHONPATH": "/path/to/Lighthouse/src"}
    }
  }
}
Air-gapped capable No outbound calls Restricted & Lockdown modes

Full install guide →

What every agent gets

69 verifier tools

Chemistry, physics, mathematics, statistics (12 tests), CS termination, biology, governance, and 50+ more — each returning CONFIRMED, MISMATCH, ERROR, or NOT_APPLICABLE. Never raises. Never self-confirms.

Full reference →

9 pipeline tools

validate_packet, seal_packet, walkthrough_packet, attest_red, attest_floor, resolve_scripture_ref, word_study, triangulate_claim, get_example_packet.

Agent guide →

4 read & synthesize surfaces

atlas walks the structural map. almanac reads the curated book of pre-run polymathic protocols and verified folk wisdom. run_polymathic coordinates every applicable verifier in parallel for cross-domain situations. propose_almanac_entry closes the curation loop.

Read the almanac →

Live data surface

22,000+ sealed packets in the journal. Append-only audit chain with SHA-256 chain integrity. Per-domain trust index. Grid scaffold across 7 axes. Voice synthesis on every narrative page (M.R. Harris).

See the grid →

"The keeping is the substrate."

The Concordance Engine at narrowhighway.com serves Jesus Christ. A well of knowledge leads to wisdom when in alignment with God. Agents are welcome to call this engine — the engine returns deterministic outcomes against external standards. It does not generate answers or speak as authority. It categorizes, it surfaces precedent, it keeps. Read the elimination trail; the trail is the reasoning.

GET /identity · /llms.txt · For AI agents