pickadive-mcp
Official PickADive MCP server — live dive-conditions, seasonality, and site data for AI agents. Free, read-only, hosted at pickadive.com/mcp
Documentation
What can you do with it?
- "Where can I dive near Boston this Saturday?" — ranked favorable and
moderate forecast windows near any place for any day, with wave height,
wind, timing, and site links.
- "When should I book Komodo?" — month-by-month climatology for a
destination: the share of days historically favorable, typical water
temperature and seas, and marine-life season windows.
- "When and where do whale sharks peak?" — species seasonality from
236k+ recorded GBIF sightings: peak month and regional season windows.
- "How deep is Molasses Reef, and what are the hazards?" — public facts
for any named dive site: depths, difficulty, shore/boat access, hazards,
entry and exit notes.
- "Where in the world is good for diving right now?" — regions ranked
by favorable windows over the next ~72 hours.
Setup
Add the server to any MCP-compatible client — no key, no OAuth:
{
"mcpServers": {
"pickadive": { "url": "https://pickadive.com/mcp" }
}
}| Client | How |
|---|---|
| Claude Code | `claude mcp add --transport http pickadive https://pickadive.com/mcp` |
| Claude.ai / Desktop | Settings → Connectors → Add custom connector → `https://pickadive.com/mcp` |
| Cursor · Windsurf · VS Code · Zed | Add the JSON block above to your MCP config |
| Gemini CLI | `gemini mcp add --transport http pickadive https://pickadive.com/mcp` |
| Local-only clients | Bridge with `npx mcp-remote https://pickadive.com/mcp` |
Discovery manifest: `/.well-known/mcp.json` ·
Agent onboarding: `/llms.txt`
Tools
| Tool | Input | Output | |
|---|---|---|---|
| `find_dive_windows` | `{near \ | lat+lng, day?, radius_km?}` | Ranked windows: quality, `wave_m`, `wind_kn`, timing, site links |
| `best_time_to_dive` | `{destination}` | 12-month climatology + marine-life seasons + source page | |
| `species_seasonality` | `{species, lat?, lng?}` | Peak month, regional season windows, sighting counts | |
| `site_facts` | `{site}` | Depths, difficulty, access, hazards, entry/exit, cert level | |
| `top_dive_regions` | `{}` | World regions ranked by favorable windows (~72 h) |
All values are metric with unit-suffixed field names (`max_depth_m`,
`wave_m`, `distance_km`); wind is knots — the marine convention
everywhere. Tool descriptions instruct agents to convert for US users.
Every response includes a source link — please cite pickadive.com
when presenting results.
Data & honesty
Answers come from PickADive's own pipeline: hourly conditions gradings
across 4,800+ curated dive sites (Open-Meteo marine + weather, NOAA
tides), multi-year climatology, and 236k+ GBIF species sighting records.
Two rules the server never breaks:
1. Forecasts are planning context, never a safety clearance. Every
response carries the disclaimer; agents should keep it.
2. Missing data is stated, never invented. Thin destinations return an
honest "not enough history" rather than a guess.
Security
| Read-only by design | No write tools — the server cannot create, modify, or delete anything. The worst a prompt injection can achieve through it is reading public dive data. |
|---|---|
| No auth, no tokens, no user data | Nothing to leak, phish, or exfiltrate; connecting grants an agent no identity. |
| Stateless JSON mode | One POST in, one JSON body out; no sessions, no server-push channel. |
| HTTPS only, CORS open | Read-only public data. |
Transport & limits
Streamable HTTP (JSON-RPC 2.0 over POST), stateless — protocol versions
`2024-11-05` → `2025-06-18`.
Free for individual and assistant use at fair, non-bulk volumes; per-IP
rate limiting applies. For sustained high-volume or commercial
integrations, write to team@pickadive.com.
Example agent guidance (optional)
Add to your `AGENTS.md` / `CLAUDE.md` to reduce round-trips:
## PickADive MCP
- For "can I dive…?" questions use find_dive_windows with the user's
place; default day is Saturday.
- For trip planning use best_time_to_dive per destination; present
depths/waves in the user's customary units, keep wind in knots.
- Always keep the safety disclaimer and cite the linked page.Support & feedback
- Issues, requests, corrections: team@pickadive.com
- Documentation: pickadive.com/developers
- The dataset behind the tools: pickadive.com/research · pickadive.com/trust
License
MIT — see LICENSE. The PickADive name and logo are trademarks
of PickADive and are not covered by the code license.
Disclaimer
Dive conditions are graded forecasts — planning context only, never a
safety judgment about any specific dive on any specific day. Always
assess conditions at the water, dive within your training, and follow
local guidance. Wildlife seasonality reflects recorded sightings, never
a promise of encounters.
Frequently asked questions
What is pickadive-mcp?
pickadive-mcp is Official PickADive MCP server — live dive-conditions, seasonality, and site data for AI agents. Free, read-only, hosted at pickadive.com/mcp
How do I install pickadive-mcp?
Open the GitHub repository and follow its README. Most MCP servers are added to your client's MCP config, then called by your agent.
Is pickadive-mcp open source?
Yes — it is hosted on GitHub at https://github.com/afiliptsov/pickadive-mcp.
Related MCP tools
🔥 Official Firecrawl MCP Server - Adds powerful web scraping and search to Cursor, Claude and any other LLM clients.
Official remote MCP server for Atlassian. Securely connect Jira, Confluence, Jira Service Management, Bitbucket, and Compass to Claude, ChatGPT, Cursor, VS Code, and other AI tools using OAuth 2.1 or API tokens.
🔥 Official Firecrawl MCP Server - Adds powerful web scraping and search to Cursor, Claude and any other LLM clients. JavaScript-based implementation.
This is a 12306 ticket search server based on the Model Context Protocol (MCP).
A model context protocol server to work with JetBrains IDEs: IntelliJ, PyCharm, WebStorm, etc. Also, works with Android Studio
A Model Context Protocol (MCP) server and CLI that provides tools for agent use when working on iOS and macOS projects.
Run your own MCP server? See who uses it and what to fix.
Measure it with TrackMCP