Neshan Maps

neshan-mcp: Neshan map of Iran MCP server for AI agents

Let your AI agent search places, get directions and check Tehran's traffic zones on Neshan (نشان).

neshan-mcp is an unofficial, read-only MCP server for Neshan, the Iranian map and navigation app.

Its 11 tools cover place search, place details, reviews, photos and menus, nearby places, reverse geocoding, city air quality, car, motorcycle, bicycle, walking and bus/metro directions, travel times to several places and Tehran traffic-zone checks. No API key or account is needed.

For AI agents · Quick start · What it can do · Tools · FAQ · فارسی

For AI agents

• What: unofficial MCP server (model context protocol) for neshan.org/maps, the Neshan map of Iran. Queries can be Persian or English; answers keep Neshan's Persian names, addresses and steps.
• Install: uvx neshan-mcp or claude mcp add neshan -- uvx neshan-mcp. Transport: stdio. Needs uv (Python 3.10+).
• Auth: none, no API key or account. Optional NESHAN_MCP_PROXY (HTTP proxy URL, or a comma-separated list); system proxy settings are ignored.
• Access: read-only. It cannot log in, save places or post reviews, so no confirmation is needed before calling any tool.
• Tools: 11, all prefixed ns_. Typical sequences: ns_search -> ns_place / ns_reviews / ns_photos / ns_nearby; ns_search -> ns_travel_times -> ns_route; ns_zone_check before driving into central Tehran.
• Use when: the user asks where something is in Iran, what is near a place, whether a business is open, what people say about it or what its menu looks like, what address a coordinate is, how polluted a city's air is, whether an address is inside Tehran's traffic zone, or how to get from A to B and how long it takes, now or at a given time.
• Do not use when: the user wants places outside Iran, taxi or delivery prices, or to save places in a Neshan account.
• Output: compact JSON. Coordinates are lat,lng (latitude first). Distances in metres or km, times in minutes, fares in Toman, review dates in the Persian calendar, departure times in Tehran time.
• Network: every tool works from an Iranian connection. Through a foreign connection all tools worked in testing except ns_nearby, which Neshan answers with HTTP 403; set NESHAN_MCP_PROXY to an Iranian proxy for it.
• Machine-readable: server.json and the iran-mcp catalog.

Why

Neshan knows Iran in detail: businesses with opening hours and reviews, live traffic, Tehran's traffic and odd-even plans, bus, BRT and metro lines. Its developer API needs a paid key, and its web map is built for clicking. neshan-mcp gives an agent the same public data the web map shows any visitor, in a few small calls:

You: How long does it take to drive from Azadi Tower to the Tehran Grand Bazaar, and will I enter the traffic zone?

Agent: *calls* ns_search("برج آزادی"), ns_search("بازار بزرگ تهران"), then

ns_route(origin="35.699739,51.338097", destination="35.671897,51.422428", steps=false)

Neshan suggests 48 minutes (13 km) via Jomhouri-e Eslami and Mostafa Khomeini streets; the route via Saeedi

expressway and Shoush is 16 km and takes 45 minutes. All three routes enter the traffic-plan zone (طرح ترافیک)

and the odd-even zone, which run from 06:30 to 18:00. The bazaar itself is inside both zones.

Real answer from a live call on 2026-10-31 at 17:21 Tehran time; times change with traffic.

Every reply is compact JSON with only the fields that matter, so it does not flood the agent's context.

What it can do

• Find places by name or category near a city or a point: cafes, pharmacies, petrol stations, streets, squares. Filter by open now and rating, the way the site does
• Read a place: address, coordinates, phones, website, rating and votes, open now, hours for each weekday, newest photos
• Read reviews, newest first or the ones Neshan features, and browse photos, including the menu album
• See what is nearby: restaurants, cafes, parks, toilets, banks, ATMs, parking and bus stops around a place
• Reverse geocode a coordinate into a street address and province
• Check air quality: today's AQI of a city, its pollutants and every monitoring station, or the index at a point
• Get directions by car, motorcycle, bicycle or on foot: alternatives, live travel time, Persian turn-by-turn steps, up to 3 stops
• Plan by time: leave at a given time or arrive by one (car, Tehran, Karaj and Mashhad), with Neshan's traffic forecast
• Rank by travel time: one origin to up to 10 places, fastest first
• Check Tehran's zones: is a point inside the traffic-plan zone or the odd-even zone, and does a route enter them; avoid them or toll roads
• Take public transport: bus, BRT and metro options with lines, stops, walking and fares in Toman
• Read-only by design: no login, no saving, no posting

Example prompts:

• "Find 24-hour pharmacies near Milad Tower that are open now and rated 4 or more."
• "Is Lamiz Coffee in Gisha open now? What's its phone number, and what do people say about it?"
• "Show me the menu photos of Milad Tower."
• "Which of these three branches is closest by car from Azadi Square?"
• "I need to be at Milad Tower at 9 tomorrow. When should I leave from Azadi Square?"
• "Is Ferdowsi Square inside Tehran's traffic zone?"
• "How polluted is the air in Isfahan today?"
• نزدیکترین داروخانه شبانهروزی به برج میلاد کجاست؟
• از میدان آزادی تا بازار تهران با ماشین چقدر راه است؟ وارد طرح ترافیک میشوم؟
• با مترو و اتوبوس از میدان آزادی به برج میلاد چطور بروم و هزینهاش چقدر است؟

Quick start

You need uv. No API key or account.

Claude Code

claude mcp add neshan -- uvx neshan-mcp

Claude Desktop

Settings → Developer → Edit Config, then add:

{

"mcpServers": { "neshan": { "command": "uvx", "args": ["neshan-mcp"] } } }

</details>

Cursor

Add the Claude Desktop block above to ~/.cursor/mcp.json (or .cursor/mcp.json in a project).

VS Code (Copilot agent mode)

Click Install in VS Code above, or add to .vscode/mcp.json:

{

"servers": { "neshan": { "type": "stdio", "command": "uvx", "args": ["neshan-mcp"] } } }

</details>

Codex CLI

codex mcp add neshan -- uvx neshan-mcp

Or in ~/.codex/config.toml:

[mcp_servers.neshan]

command = "uvx" args = ["neshan-mcp"]

</details>

Gemini CLI

Add to ~/.gemini/settings.json:

{

"mcpServers": { "neshan": { "command": "uvx", "args": ["neshan-mcp"] } } }

</details>

Windsurf

Add the Claude Desktop block to ~/.codeium/windsurf/mcp_config.json.

Anything else

It is a standard stdio MCP server: run uvx neshan-mcp, or pip install neshan-mcp and run neshan-mcp.

Tools

Points are lat,lng, e.g. 35.6997,51.3380. Place ids come from ns_search.

All 11 tools are annotated readOnlyHint: true.

How it works

  AI agent  (Claude, Cursor, Copilot, ...)

│ │ MCP over stdio ▼ neshan-mcp (runs on your machine) │ │ HTTPS └──────▶ neshan.org/maps/pwa-api search, places, reviews, photos, routes, transit, air quality

neshan-mcp runs locally and calls the same endpoints the neshan.org/maps web app uses, with no hosted server in between and no API key. Requests go out through a Chrome-impersonating client and look like the web app's own. It sends at most 3 requests at a time with a little jitter, caches answers (places and addresses for 5 minutes, live routes for 1 minute, transit, zones and air quality for 10 minutes), retries a failed call up to twice, and only talks to neshan.org.

Neshan's web app encodes most of its answers: they arrive as shuffled base64 text, and transit answers are also XOR-encoded. The server decodes them the same way the web app's own script does. If Neshan changes the scheme, the server reads the new values from the site's current JavaScript once, the way a browser loads it, and retries. An answer it still cannot read is reported as an error, never as made-up data.

Good to know

• Coordinates are `lat,lng`. Latitude first. A swapped pair is rejected with a hint.
• `near` takes a city or a point. 23 big cities by name in English or Persian (Tehran, Mashhad, Isfahan, Shiraz, Tabriz, Karaj, Qom, Ahvaz, ...). For a smaller town, put its name in the query.
• Search returns at most 30 places, with no paging. Sponsored results (marked آگهی on the site) are left out. open_now and min_rating use Neshan's own filters (open now; 3.5+, 4