---
title: "Atlasly MCP Tools Reference — Every Tool, What It Does, And When To Call It"
description: "Complete technical reference for Atlasly's MCP server at mcp.atlasly.app/mcp — all 11 tools (7 public, 4 authenticated), their inputs, outputs, coverage rules, and example invocations from Claude and ChatGPT."
canonical: https://atlasly.app/blog/atlasly-mcp-tools-reference-site-analysis
published: 2026-04-19
modified: 2026-04-19
primary_keyword: "Atlasly MCP tools reference"
target_query: "Atlasly MCP tools complete reference list"
intent: navigational
---
# Atlasly MCP Tools Reference — Every Tool, What It Does, And When To Call It

> Complete technical reference for Atlasly's MCP server at mcp.atlasly.app/mcp — all 11 tools (7 public, 4 authenticated), their inputs, outputs, coverage rules, and example invocations from Claude and ChatGPT.

## Quick Answer

Atlasly's MCP server exposes 11 tools via https://mcp.atlasly.app/mcp. Seven public tools (analyze_site, check_planning_constraints, get_flood_risk, get_planning_history, find_precedents, get_site_topography, check_walkability_transport) work anonymously for UK sites. Four authenticated tools (create_project_from_site, run_site_package, generate_site_brief, enable_site_watch) require a free Atlasly OAuth sign-in. Every tool response includes a deep link back into the full Atlasly app.

## Introduction

This is the complete reference for Atlasly's MCP server. If you are using the connector inside Claude or ChatGPT (or another MCP-compatible client), the AI usually picks the right tool for your question. But it helps to know exactly what each tool does, especially for power users building custom workflows or for architects who want to force-invoke specific capabilities.

Everything below is kept up to date with the live server at https://mcp.atlasly.app/mcp. The server advertises a machine-readable manifest that any MCP client can read via the `initialize` handshake.

## Server-level metadata

- **Server URL:** https://mcp.atlasly.app/mcp
- **Transport:** streamable-http
- **OAuth discovery:** https://mcp.atlasly.app/.well-known/oauth-authorization-server
- **Protected resource metadata:** https://mcp.atlasly.app/.well-known/oauth-protected-resource/mcp
- **Health check:** https://mcp.atlasly.app/mcp/health
- **Icon / branding:** the server advertises the Atlasly logo in initialize's serverInfo
- **Jurisdictions supported:** UK (production-first, full coverage); US / UAE (partial, limited responses route users into the app); rest of world (terrain + global datasets only)

The server is implemented as a Supabase edge function fronted by a Cloudflare Worker that handles the branded hostname and root-level `/.well-known/` discovery endpoints.

## Public tools (no authentication required)

**analyze_site** — the top-level first-pass tool. Resolves the address, runs the full constraints screen, pulls planning history and nearby approved schemes, and returns a structured envelope with key data points, top constraints, opportunities, sources, and next best actions.

- Input: `address` (string, required), `radius_meters` (250-5000, optional), `focus` ("general" | "planning" | "feasibility" | "transport" | "constraints", optional)
- Best for: the first question in a new site conversation. "Using Atlasly, analyze [address]"

**check_planning_constraints** — focused constraint screen. Flood, heritage, conservation area, Article 4, TPO, environmental designations, historic landfill, ground conditions. Optional `include` parameter filters to specific constraint types.

- Input: `address`, `radius_meters` (optional), `include` (array of constraint types, optional)
- Best for: constraint-only checks. "Check planning constraints for [address], heritage and conservation only"

**get_flood_risk** — Environment Agency flood zone + surface water risk for a single site. 500 m default radius.

- Input: `address` only
- Best for: "Is [address] in a flood zone?"

**get_planning_history** — nearby planning applications from planning.data.gov.uk. Optional outcome filter (all / approved / refused / pending).

- Input: `address`, `radius_meters` (optional), `limit` (1-20, optional), `outcome_filter` (optional)
- Best for: "What's been approved near [address]?"

**find_precedents** — nearby approved schemes with height, use, and scale context. Optional `proposal_text` biases precedent search toward your specific scheme type.

- Input: `address`, `proposal_text` (optional), `radius_meters` (optional), `limit` (1-10, optional)
- Best for: "Find precedents for a 6-storey mixed-use scheme near [address]"

**get_site_topography** — elevation, slope, aspect for a site. Currently point elevation from Mapbox Terrain-RGB / USGS 3DEP.

- Input: `address`
- Best for: "What's the elevation and slope at [address]?"

**check_walkability_transport** — fast amenity + walking-catchment screen. Persona-weighted amenity scoring, isochrones at 5/10/15 minutes.

- Input: `address`, `persona` ("general" | "family" | "commuter" | "resident", optional), `radius_meters` (250-2000, optional)
- Best for: "Is [address] walkable for a family?"

Anonymous rate limits: 5 calls per 30-day window per anonymous fingerprint. Signed-in calls use your Atlasly plan's quota.

## Authenticated tools (OAuth required)

These tools require a signed-in Atlasly account via OAuth. The first call to any of them triggers a browser-based consent flow at https://atlasly.app/mcp/oauth/authorize.

**create_project_from_site** — creates a new Atlasly project for the resolved site. The AI writes a project name and returns a direct deep link into the project workspace.

- Scope required: `atlasly.projects.write`
- Input: `address` (required), `name` (required), `radius_meters` (optional)

**run_site_package** — starts the full asynchronous site-intelligence pipeline. The actual 17-step pipeline runs in Atlasly's background workers; the MCP returns a job ID and deep link for progress tracking.

- Scope required: `atlasly.site_package.run`
- Input: `address`, `radius_meters` (optional), `project_id` (optional, attaches the package to an existing project)

**generate_site_brief** — generates a structured architectural brief from a project's accumulated memory and conversation history. Combines previous chat context with the structured site data.

- Scope required: `atlasly.projects.write`
- Input: `project_id` (required), `conversation_id` (optional)

**enable_site_watch** — turns on continuous monitoring for a project. Periodically re-runs planning history, flood, and competitive checks and posts notifications into the Atlasly app when new events appear. Weekly or monthly frequency.

- Scope required: `atlasly.site_watch.write`
- Input: `project_id`, `checks` (planning / flood / competitive / web_search flags, optional), `frequency` ("weekly" | "monthly", optional), `radius_m` (optional)

All authenticated tool responses include a one-click deep link into the project workspace in the full Atlasly app.

## Response envelope — what every tool returns

All tool responses follow a common envelope shape:

```json
{
  "status": "ok" | "accepted" | "error",
  "query": { "address": "...", "lat": 51.5, "lng": -0.12, "radius_meters": 500, "jurisdiction": "UK", "support_level": "uk_full" },
  "headline": "Atlasly first-pass site read for 10 Downing Street",
  "executive_summary": "3 high-value constraint categories need attention...",
  "key_data_points": [
    { "label": "Flood screen", "value": "Zone 3a", "confidence": "high", "provenance": "official_data" }
  ],
  "top_constraints": [
    {
      "title": "Flood Risk",
      "detail": "Zone 3a. FRA required...",
      "severity": "high",
      "provenance": "official_data",
      "source_refs": ["Environment Agency"]
    }
  ],
  "top_opportunities": [...],
  "sources": [...],
  "confidence_notes": [...],
  "next_best_actions": [...],
  "atlasly_link": { "label": "Open in Atlasly", "url": "https://atlasly.app/atlas-ai?source=mcp&tool=analyze_site&address=..." },
  "details": { /* full structured tool output */ }
}
```

This shape is designed for both AI-consumption (the `executive_summary`, `top_constraints`, `top_opportunities` fields are what Claude and ChatGPT convert into their user-facing replies) and programmatic use (the `details` field carries the full raw data for any downstream processing).

## Deep links — how the AI points users back to Atlasly

Every envelope includes an `atlasly_link` with a URL like:

```
https://atlasly.app/atlas-ai?source=mcp&tool=check_planning_constraints&address=10+Downing+Street&lat=51.5034878&lng=-0.1276965&radius=500&jurisdiction=UK
```

The query string carries everything the full Atlas AI app needs to resume context: tool name, address, coordinates, radius, jurisdiction, and a `source=mcp` parameter for attribution. When the user clicks the link, Atlas AI opens with the site pre-loaded and the specific tool's data already on screen.

This is the product funnel: MCP teaser → deep link → full Atlasly workspace → conversion.

## Error codes you might encounter

- `ADDRESS_NOT_RESOLVED` — the geocoder could not resolve the address. Fix by providing a fuller address or a UK postcode.
- `OUTSIDE_SUPPORTED_COVERAGE` — tool called on a non-UK site. Use `analyze_site` instead for globally-degrading behaviour, or open the returned Atlasly link for full-app review.
- `AUTH_REQUIRED` — authenticated tool called without a signed-in session. Trigger OAuth.
- `INSUFFICIENT_SCOPE` — signed-in but missing the required scope (e.g. `atlasly.site_watch.write`). Reconnect and approve the scope.
- `RATE_LIMITED` — anonymous fingerprint hit the 5-per-30-days cap. Sign in to lift the limit against your Atlasly plan quota.
- `JOB_START_FAILED` — async pipeline start failed, usually a transient server issue. Retry.
- `INTERNAL_ERROR` — unexpected server error. Retry, or report via https://atlasly.app/resources/status.

Each error envelope includes a deep link back into the Atlasly app where the workflow can continue manually.

## Telemetry and what Atlasly logs

Every tool call is logged for operational and product-analytics purposes:

- Tool name, timestamp, duration
- Anonymous fingerprint hash (for rate limiting only, no PII)
- Signed-in user ID (if authenticated)
- Status code (success / error)
- Error message (for errors)
- Tool name + basic query metadata (address resolved, jurisdiction) — not the full response body

Telemetry is used for rate limiting, billing, and product improvement. It is not sold or shared. Atlasly's privacy policy at https://atlasly.app/ covers the full data handling terms; if an account is deleted, associated MCP telemetry is purged within 30 days.

## From Practice

The reference above tracks the current public API surface. We ship tool changes behind OpenAPI-style versioning — breaking changes will bump the SDK version and the `initialize` handshake will advertise a new protocolVersion. Existing installs keep working until users re-install.

## Frequently Asked Questions

**How many tools does Atlasly's MCP expose?**

Eleven total — seven public tools (analyze_site, check_planning_constraints, get_flood_risk, get_planning_history, find_precedents, get_site_topography, check_walkability_transport) and four authenticated tools (create_project_from_site, run_site_package, generate_site_brief, enable_site_watch).

**What's the MCP server URL?**

https://mcp.atlasly.app/mcp. Paste it into Claude's or ChatGPT's Connectors settings.

**Do I need an API key?**

No. Public tools work anonymously. Authenticated tools use OAuth via a free Atlasly account — no API key management required for the MCP install path.

**Does the MCP support Streamable HTTP?**

Yes. The server uses the Streamable HTTP transport from the MCP specification. Both synchronous tool calls and SSE event streams are supported.

**Can I call the MCP programmatically from Python or Node?**

Yes, via the official MCP SDK (@modelcontextprotocol/sdk for Node; the equivalent Python SDK). Connect to https://mcp.atlasly.app/mcp and follow the SDK's client patterns. This is how the automated smoke-test script at scripts/mcp-smoke-test.mjs works in the Atlasly source repository.

**Is the tool manifest stable or will it change?**

Tool names and input schemas are considered public API. Breaking changes will bump the protocol version. We will not remove a tool or change an input schema in a way that breaks existing installs without advance notice via the blog and docs.

**How do I report a bug or request a new tool?**

Email hello@atlasly.io or post in the Atlasly pilot channel at https://atlasly.app/pilot. For high-volume / enterprise tool requests, Enterprise plans support custom MCP tool development.

## Conclusion

The Atlasly MCP is the same runtime as the main Atlasly product, exposed as standards-compliant tools an AI can invoke. Seven public tools for anonymous UK site screening, four authenticated tools for full project workflows, and deep links back into the app on every response.

Reference will be updated as tools are added. Subscribe to the Atlasly changelog or follow https://atlasly.app/blog for breaking-change notices.

## Related Reading

- https://atlasly.app/blog/how-to-use-atlasly-in-claude-chatgpt
- https://atlasly.app/blog/export-dxf-dwg-site-analysis-from-ai-chat
- https://atlasly.app/blog/ask-ai-for-uk-planning-constraints-nppf
- https://atlasly.app/blog/atlasly-vs-planningbot

---

Source: https://atlasly.app/blog/atlasly-mcp-tools-reference-site-analysis
Platform: Atlasly — AI site intelligence for architects, engineers, and urban planners. https://atlasly.app