Overview

AI-generated overview

Lets an assistant scan Street View areas into PanoIDs and download the imagery through a locally running PanoOn service.

What it does: Wraps the PanoOn local service and distributed controller over stdio. Tools cover service status, area scans by polygon, scan status, results and cancellation, plus imagery downloads, job status, job listing, worker listing and job cancellation. Scans return PanoIDs; downloads write image files to a chosen output directory.
When to use it: Use it when you want an MCP-capable coding agent to collect Street View coverage for a geographic area and fetch the corresponding panoramas, with job progress and cancellation handled through the agent.
Requirements: A local process started with Node.js via npx panoon-mcp, or the runtime bundled with PanoOn. The PanoOn app or service must be running on the same machine; PANOON_SERVICE_URL sets its base URL (default Desktop only.

Before you install

Downloads are large (roughly 0.5 MB per image) and consume proxy quota, so confirm scope and output paths before submitting jobs. Download tools write files to a user-specified directory. Scans are described as cheap and cancellable.

Installation

In SourceWeft

Open PanoOn Street View in the dashboard and add it to a workspace.
Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

PanoOn MCP + agent skills

Give any MCP-capable code agent the ability to scan and download Street View data through PanoOn: mcp/panoon-mcp.mjs is a zero-dependency MCP stdio server that wraps the PanoOn local service and the distributed controller.

Quick start (npm)

The npm package panoon-mcp exposes the panoon-mcp bin. You can also download panoon-mcp.zip from https://app.rjfan.org/panoon/mcp/, or use the copy bundled with PanoOn 0.2.21+ (<install>\mcp\panoon-mcp.mjs).

Prerequisite: the PanoOn app/service is running on the same machine (http://127.0.0.1:8787; the desktop app starts it automatically since 0.2.20).

The server

sh

node mcp/panoon-mcp.mjs# or with the runtime bundled with PanoOn (no Node install required):"C:\Users\<you>\AppData\Local\PanoOn\panoon-service-runtime.exe" mcp\panoon-mcp.mjs

node mcp/panoon-mcp.mjs# or with the runtime bundled with PanoOn (no Node install required):"C:\Users\<you>\AppData\Local\PanoOn\panoon-service-runtime.exe" mcp\panoon-mcp.mjs

Environment:

Tools: panoon_status, panoon_area_scan, panoon_area_status, panoon_area_result, panoon_area_cancel, panoon_download, panoon_job_status, panoon_jobs, panoon_workers, panoon_cancel_job.

Install per agent

OpenCode (V2)

or by hand in ~/.config/opencode/opencode.jsonc:

jsonc

{  "mcp": {    "servers": {      "panoon": {        "type": "local",        "command": ["node", "<repo>/mcp/panoon-mcp.mjs"],        "environment": { "PANOON_SERVICE_URL": "http://127.0.0.1:8787" }      }    }  }}

{  "mcp": {    "servers": {      "panoon": {        "type": "local",        "command": ["node", "<repo>/mcp/panoon-mcp.mjs"],        "environment": { "PANOON_SERVICE_URL": "http://127.0.0.1:8787" }      }    }  }}

Skill: copy mcp/skills/panoon/ into your OpenCode skills location (or reference the folder from the skills configuration field).

Claude Code

or .mcp.json:

Skill: copy mcp/skills/panoon/ to ~/.claude/skills/panoon/.

Codex CLI

Add to ~/.codex/config.toml:

toml

[mcp_servers.panoon]command = "node"args = ["<repo>/mcp/panoon-mcp.mjs"]env = { PANOON_SERVICE_URL = "http://127.0.0.1:8787" }

[mcp_servers.panoon]command = "node"args = ["<repo>/mcp/panoon-mcp.mjs"]env = { PANOON_SERVICE_URL = "http://127.0.0.1:8787" }

For Codex, paste the workflow section of mcp/skills/panoon/SKILL.md into your AGENTS.md so the agent knows the intended flow.

Any other MCP client (stdio)

Use the generic stdio contract:

json

{  "command": "node",  "args": ["<repo>/mcp/panoon-mcp.mjs"],  "env": { "PANOON_SERVICE_URL": "http://127.0.0.1:8787" }}

{  "command": "node",  "args": ["<repo>/mcp/panoon-mcp.mjs"],  "env": { "PANOON_SERVICE_URL": "http://127.0.0.1:8787" }}

DeepSeek Harness (dsh)

This writes ~/.deepseek/mcp.json (servers.panoon; tools surface namespaced by the harness). Install the skill by copying mcp/skills/panoon/ into ~/.deepseek/skills/panoon/ — dsh uses the same SKILL.md contract as Claude Code — then restart the harness so both are picked up.

On this machine the working example is:

sh

deepseek mcp add panoon --command node --arg D:\04_Dev\PanoOn\mcp\panoon-mcp.mjscopy D:\04_Dev\PanoOn\mcp\skills\panoon %USERPROFILE%\.deepseek\skills\panoon

deepseek mcp add panoon --command node --arg D:\04_Dev\PanoOn\mcp\panoon-mcp.mjscopy D:\04_Dev\PanoOn\mcp\skills\panoon %USERPROFILE%\.deepseek\skills\panoon

Typical agent flow

panoon_status → service healthy, strategies available.
panoon_area_scan { polygon: [{lat,lng}, ...] } → jobId; panoon_area_result → PanoIDs.
Confirm output directory with the user.
panoon_download { panoids, outputDir } → jobId; panoon_job_status for progress.

Safety: downloads are large (~0.5 MB/image) and consume proxy quota. Agents should confirm scope and output paths before submitting jobs; scans are cheap and can be cancelled at any time.

Source: README.md at commit 6aa8e42

Tools

0

Tool metadata has not been indexed yet.

Version history

1

v0.2.21LatestOct 7, 2026

PanoOn Street View