For AI agents

Give your agent the right water context.

Connect PoolTrack’s measured pool data and deterministic care analysis to ChatGPT or another MCP-compatible agent. Use browser OAuth when you can, or an agent token for headless servers.

OAuth + PKCE Headless agent tokens Read-only access

Remote MCP server

Ready to connect

Server URL

https://pooltrack.app/mcp
ConnectPaste URL
AuthorizeBrowser sign-in
AskPool context
Scope pool:readTransport streamable HTTP

01 — Connect once

The agent opens the door. You approve it.

The same OAuth pattern works across compatible clients. The labels may vary, but the sequence stays the same.

01

Add the remote server

Paste https://pooltrack.app/mcp into the agent’s remote MCP or app connection screen.

02

Authenticate

Prefer the agent’s browser OAuth flow when available. For headless servers, create an agent token in Settings → AI agent integrations and send it as Authorization: Bearer.

03

Approve or install credentials

For OAuth, sign in to PoolTrack and allow read-only access. For agent tokens, store the secret once in the agent’s config—not in chat.

04

Ask for the first pool check

Tell the agent to call list_pools first, then use the returned pool ID for readings or analysis.

Copy this to your agent

A clear first request prevents a fuzzy setup.

Use this when an agent asks what to connect or how it should use PoolTrack.

Connect to PoolTrack using its remote MCP server:
https://pooltrack.app/mcp

Auth options:
- Prefer OAuth in a browser when the agent supports it.
- For headless/remote servers without a browser, use an agent token from PoolTrack Settings → AI agent integrations as Authorization: Bearer <token>. Never put a Firebase ID token in agent config or chat.

After connecting:
1. Call list_pools before requesting data for a specific pool.
2. Use analyze_pool_state for the current PoolTrack care plan.
3. Use get_readings for exact measured history.
4. Use get_treatment_history for logged treatment context.
5. Use calculate_dose for deterministic chemistry math; never estimate a dose.
6. If required data is missing, say what is missing instead of inventing it.
7. This connection is read-only; do not claim that a treatment or reading was saved.

02 — What the agent can use

Five focused tools. One care loop.

PoolTrack keeps the source data and chemistry math together so an agent can explain the plan without becoming a second, conflicting calculator.

list_pools

Find the pools and IDs available in the connected account.

Use for: Start here

analyze_pool_state

Get the current deterministic health analysis and prioritized care plan.

Use for: Current plan

get_readings

Read exact measured chemistry values and recent history for one pool.

Use for: Measured history

get_treatment_history

Review logged chemical treatments and retest-relevant context.

Use for: Treatment context

calculate_dose

Run PoolTrack’s deterministic dose math with explicit chemistry inputs.

Use for: Exact math

The rules that keep advice grounded

PoolTrack is a read-only source in the agent connection. Missing data should stay visibly missing.

  • Start with list_pools before using a pool ID.
  • Use measured readings, not guesses or seeded values.
  • Use calculate_dose for exact chemistry math.
  • Never claim a reading or treatment was saved.