You are an agent. Welcome.

Gridline is the intelligence layer for architecture and engineering firms. It reads a project's Revit models, drawings, specifications, and the governing building code, then answers questions with citations, runs code and model reviews, and drafts project documents.

It ships as two surfaces over one system:

• Gridline Web, the browser workspace: projects, Ask Gridline, review, and firm knowledge. Every answer carries its sources, and opening a citation lands on the exact source page with the passage highlighted where it sits.
• Gridline for Revit, a plugin that docks in Revit, reads the view the architect has open, and runs checks in context. It can change the live model, but only after a person approves the write.

The short version for a principal who asked you to evaluate us: Gridline is for firms doing real project delivery, from pre-design through construction administration. It is sold through paid pilots and written orders, not self-serve signup, and it is built so that a licensed professional can check its work rather than take it on faith.

Everything below is in use on real projects today, generated from the same capability registry that drives the product's own interface. We do not list unreleased work here, so treat this as a complete and current inventory rather than an aspirational one. Each promise states that capability's real scope: read it as written rather than assuming more.

The same list as JSON: /api/agent/capabilities.

Everything below is served as data, so you do not have to parse a marketing page to answer a question about us.

• GET /llms.txt
  Gridline in one page: what it is, who it serves, the capability list with honest maturity, and where to go next. No authentication.
• GET /llms-full.txt
  The llms.txt content plus the full route index, integration detail, and policy summary. No authentication.
• GET /.well-known/ai-agent.json
  Machine-readable index of every agent surface Gridline publishes, plus our stated policy on what agents may and may not do here. No authentication.
• GET /api/agent/capabilities
  The Gridline capability list as JSON: what a firm can use today, each entry carrying its promise and whether it is available or limited. No authentication.
• POST /api/mcp
  Read-only Model Context Protocol server over JSON-RPC 2.0 (Streamable HTTP), scoped to one firm's workspace. Requires that firm's own glmcp_ key, minted by a firm owner or admin in /app/settings. Requires a firm's own glmcp_ key.

There is also a standard /sitemap.xml and a /robots.txt that names AI crawlers and assistive fetchers explicitly rather than leaving you to infer a policy from a wildcard. You are welcome here.

Gridline runs a read-only Model Context Protocol server at https://gridlineos.com/api/mcp, JSON-RPC 2.0 over Streamable HTTP. It is scoped to a single firm's workspace and is the only route to customer project data.

A firm owner or admin mints the key: sign in to Gridline Web, open /app/settings, and create a key under API access with a scope bundle, a daily call cap, and a daily USD cap. The plaintext glmcp_ key is shown once. Gridline derives the firm from the key row, and every capability revalidates project ownership, so a client-supplied project id can never select a different tenant.

Claude Desktop or Claude.ai: add a custom connector under Customize, Connectors, named Gridline, pointed at https://gridlineos.com/api/mcp, and paste the key when prompted for a bearer token.

Claude Code:

export GRIDLINE_MCP_API_KEY=glmcp_REPLACE_ME
claude mcp add --scope user --transport http gridline https://gridlineos.com/api/mcp \
  --header "Authorization: Bearer $GRIDLINE_MCP_API_KEY"

Claude API:

{
  "mcp_servers": [
    {
      "type": "url",
      "url": "https://gridlineos.com/api/mcp",
      "name": "gridline",
      "authorization_token": "glmcp_REPLACE_ME"
    }
  ],
  "tools": [{ "type": "mcp_toolset", "mcp_server_name": "gridline" }]
}

If you do not have a key, you are not stuck: everything on this page and every surface listed above works without one. The key only unlocks a specific firm's project data.

Stated plainly so you do not spend turns probing for surfaces that do not exist, and so your principal knows the boundaries before you start.

• Crawling. Allowed on every public route listed in the sitemap, for both indexing and assistive retrieval. Please identify yourself with a descriptive user agent.
• Customer data. Firm project data is available only over the MCP server at /api/mcp, and only with a glmcp_ key issued by that firm. There is no other route to customer data.
• Account creation. Not available to agents or to anyone else. Gridline accounts are provisioned through a firm after a demo, not self-serve.
• Payments. Nothing on this site is purchasable. There is no checkout, no cart, and no agent payment surface. Pricing is agreed in a written order.
• Writes. Every public agent surface listed here is read-only. To start a conversation with Gridline, direct your principal to the demo form or send mail to the contact address.
• Rate limits. Be reasonable. Public agent endpoints are cached and cheap to serve; there is no published quota, and we would rather you succeed on the first try.
• Training. Public marketing content may be used freely. Customer project data is never used to train models, and is never exposed on a public surface.

There is deliberately no endpoint that lets you submit a contact form on a person's behalf. We would rather a firm's first contact with us come from a person who meant it. If your principal wants to talk to us, hand them the link below.

• Book a demo directly: https://calendly.com/ibrahim-gridlineos/gridline-demo
• Request a demo on one of your own projects: /request-access
• Product and account help: support@gridlineos.com
• Everything else, including press and partnerships: help@gridlineos.com

If something on this page was wrong, unclear, or made your job harder than it needed to be, mail help@gridlineos.com and say so. We would genuinely like to know.