MCP Server (AI Connectors)
WildTrack360 exposes a Model Context Protocol (MCP) server so AI chat apps — Claude, ChatGPT, Cursor, Claude Code, and any other MCP-capable client — can securely read and act on your organisation's data on your behalf.
:::info Who this is for This page is for people who want to connect an AI assistant to their WildTrack360 workspace. If you only use the web app and Wally, you can safely skip it. For Wally, the assistant built directly into WildTrack360, see Wally AI Assistant. :::
What it is
An MCP server is a standard way for an outside AI client to connect to an application and use a fixed set of tools. Once you connect a client to WildTrack360, you can ask it — in plain English — to list your animals, look one up, add a care record, run a report, and more, and it carries those actions out through the MCP tools.
- Endpoint:
POST https://<your-org>.wildtrack360.com.au/mcp(Streamable HTTP transport) - Sign-in: Clerk OAuth 2.0. The client authenticates as you — every tool call is scoped to your user, your organisation, and your role and species access, exactly the same visibility rules as the web app and API. The client never gets more access than you have.
One-time setup
Because MCP clients register themselves during sign-in, an administrator has to enable that once in Clerk:
- Open the Clerk Dashboard → OAuth applications.
- Toggle on Dynamic client registration.
No other configuration is required — WildTrack360 serves the OAuth discovery endpoints itself.
Connecting a client
| Client | How to connect |
|---|---|
| Claude (claude.ai / Desktop) | Settings → Connectors → Add custom connector → URL https://<your-org>.wildtrack360.com.au/mcp. Claude walks you through the WildTrack360 sign-in and consent screen. |
| ChatGPT | Settings → Connectors (developer mode) → add an MCP server with the same URL. |
| Claude Code | claude mcp add --transport http wildtrack360 https://<your-org>.wildtrack360.com.au/mcp, then run /mcp to authenticate. |
| Cursor / other | Add a Streamable HTTP MCP server pointing at /mcp; the client discovers the sign-in automatically. |
:::tip Connect through your organisation's subdomain
Connect using your organisation's own address (for example https://yourorg.wildtrack360.com.au/mcp). The session is then pinned to that organisation, matching how the web app scopes you. If you connect through the root domain instead, tools default to your first organisation.
:::
What the AI can do
All tools reuse the same permission rules as the app. Only one tool writes data, and it is audit-logged.
| Tool | Who can use it | What it does |
|---|---|---|
whoami | Any member | Shows who you are, your organisations, and your role in the active org. |
list_animals | Role-scoped | Lists animals visible to your role. Filter by species, status, or search. |
get_animal | Role-scoped | Full detail for one animal (by internal ID or org animal ID) with its 20 most recent care records. |
create_care_record | Animals you can access | The only write. Adds a feeding, medical, behaviour, location, weight, release, or general record to an animal you can access. Audit-logged. |
list_carers | Coordinator and above | Carer roster with contact, licence, and specialty details. |
list_species | Any member | Species configured for your organisation. |
run_report_query | Coordinator and above | Runs read-only Custom Reporting queries via the same safe QL engine as the report workbench. |
list_saved_report_queries | Coordinator and above | Lists saved report queries, re-runnable through run_report_query. |
get_report_query_reference | Any member | The QL grammar plus every queryable source and its fields. |
Security
- You are never impersonated. Every request runs as your Clerk user with your RBAC role and species access. An unauthenticated request is rejected.
- Membership is always verified. The organisation address only selects which org a tool acts on; access is checked against your Clerk memberships every time.
- Reporting is read-only and safe.
run_report_querygoes through the same anchored QL parser and allowlist as the report workbench — no raw SQL is ever built from your text, and results stay scoped to your organisation. - Writes are audited. The single write tool,
create_care_record, is recorded in the audit log just like a record added in the app.
MCP vs Wally
| MCP Server | Wally | |
|---|---|---|
| Where it runs | An outside AI client you connect (Claude, ChatGPT, Cursor…) | Built into the WildTrack360 web app |
| Sign-in | Clerk OAuth, acting as you | Your existing app session |
| Writes | Add care records only | Admit animals, add care records, log training, record growth |
| Best for | Working from your preferred AI assistant | Quick actions and answers without leaving the app |
Both respect the same roles, species access, and audit trail.
Related Documentation
| Page | Why it matters |
|---|---|
| Wally AI Assistant | The in-app assistant; overlapping tools, no external client needed |
| API Reference | The REST API behind the same data, for custom integrations |
| Roles & Permissions | The RBAC and species-access rules every MCP tool respects |
| Custom Reporting QL | The query language behind run_report_query |
| Audit Logging | Where MCP writes are recorded |