Model Context Protocol (MCP)

Your compatible AI tools,
with relevant project context.

Connect supported or MCP-compatible clients to ReasonTrail. Retrieve approved, bounded project context and capture work with provenance when explicitly configured.

View connection options Download the Mac app

MODEL-AGNOSTIC CONNECTION (HTTP TRANSPORT)

Connect an MCP-compatible client
to ReasonTrail.

Live Endpoint

ReasonTrail Remote MCP

Authenticated Streamable HTTP Endpoint

Connect an MCP-compatible client using HTTP transport. Generate a scoped token in your ReasonTrail account settings under Settings → MCP clients.

Server Endpoint URLhttps://app.reasontrail.ai/api/mcp

Required Connection Parameters:

  • Transport: http (Streamable HTTP / Server-Sent Events)
  • URL: https://app.reasontrail.ai/api/mcp
  • Header: Authorization: Bearer <your-mcp-token>

1-Line CLI Connection Command:

claude mcp add --transport http reasontrail https://app.reasontrail.ai/api/mcp \
  --header "Authorization: Bearer <your-mcp-token>"

Connection & Authorization Flow

Account & Organization Scoped
  1. 01

    Sign in to ReasonTrail & open Settings → MCP clients

  2. 02

    Generate a Personal Access Token (Bearer token)

  3. 03

    Copy the token into your AI assistant or client configuration

  4. 04

    Connect using HTTP transport or JSON config

  5. 05

    Your assistant retrieves bounded, authorized context with project provenance

JSON Config Format (Cursor, Windsurf, VS Code, Zed, Claude Desktop)

{
  "mcpServers": {
    "reasontrail": {
      "url": "https://app.reasontrail.ai/api/mcp",
      "headers": {
        "Authorization": "Bearer <your-mcp-token>"
      }
    }
  }
}

ReasonTrail is the authorization boundary.

Your AI assistant can request context or propose changes. ReasonTrail enforces tenant isolation, project permissions, and safety policies.

Model-agnostic protocol for compatible MCP clients One validated active organization per token Approved tools and scoped project access Revocable access tokens with customizable expiry Membership and suspension rechecked on every request Bounded context output with credential protection No database passwords or secret keys exposed No unrestricted shell or host filesystem access

ReasonTrail does not expose raw database credentials, environment keys, or unrestricted shell/filesystem access to AI models.

ADVANCED / LOCAL MODE

Developer and
self-hosted local stdio MCP.

Run the ReasonTrail MCP server locally over stdio on your self-hosted instance or development machine.

ReasonTrail is the MCP server A compatible AI client is the MCP caller Model-agnostic context delivery
01

Prepare your local installation

Install dependencies and configure the environment for your ReasonTrail instance.

cd /path/to/reasontrail
npm install
02

Bind the server to your account email

If the installation has multiple users, set BRAIN_USER_EMAIL in your client environment so the server binds to your specific account.

03

Configure your client command

Add the stdio server command to your AI client configuration.

CLI Command (Model-Agnostic)Run in Terminal
claude mcp add reasontrail --env BRAIN_USER_EMAIL=you@example.com -- npx tsx /absolute/path/to/reasontrail/scripts/mcp-server.mts
TOML Config (Codex)~/.codex/config.toml
[mcp_servers.reasontrail]
command = "npx"
args = ["tsx", "/absolute/path/to/reasontrail/scripts/mcp-server.mts"]
cwd = "/absolute/path/to/reasontrail"
env = { BRAIN_USER_EMAIL = "you@example.com" }
Claude Desktop / Generic Stdio JSON Config Show JSON
{
  "mcpServers": {
    "reasontrail": {
      "command": "npx",
      "args": ["tsx", "/absolute/path/to/reasontrail/scripts/mcp-server.mts"],
      "env": {
        "BRAIN_USER_EMAIL": "you@example.com"
      }
    }
  }
}
04

Restart your AI assistant

Your AI client automatically launches and communicates with the ReasonTrail MCP server.

CONNECTED PROJECT MEMORY

Continue working.
Keep the reasoning.

Open ReasonTrail to review what connected AI assistants capture and propose.

Request private beta