AI at AMMP
We are putting your portfolio data directly in front of AI. An open MCP server gives any assistant secure access to your assets, KPIs, alerts and raw telemetry, and our own agents are being built on exactly the same foundation.
Open protocol
The AMMP MCP server
The AMMP MCP server speaks the open Model Context Protocol, so it plugs into Claude, Cursor, Copilot or an agent you build yourself. The model decides which tools to call, runs the queries against your live portfolio and answers in natural language.
What it exposes
Assets and metadata
Your full portfolio structure: sites, devices, capacities, commissioning data and the metadata that gives every reading context.
Computed KPIs
Production, performance ratio, availability, specific yield and the rest of the KPI set, at the same definitions your dashboards use.
Alerts, active and historical
Current alerts and the full alert history, so an assistant can correlate today's underperformance with an event three weeks ago.
Raw time-series
Telemetry at native resolution, so answers can be cross-checked against the underlying measurements rather than only the aggregates.
Permissions and scope follow your existing AMMP access controls. An assistant can only see what the connected user can already see.
How this can fit into your stack
A rough map of the integration options we are testing with beta customers. Use our agents, bring your own assistant, or both.
Provided by AMMP
Clay agent
Slack and Teams coworker
- Remembers your context
- Learns from feedback
- Combines data sources
- Builds reports and exports
Ray agent
On-demand visualisation
- Built into AMMP OS
- Interactive charts
- Live dashboards
- On-the-fly troubleshooting
Your AI assistant
Claude, Codex, Copilot
- Native chat interface
- Direct portfolio access
- Full MCP tool support
Other software tools
Optional integration
- CRM, CMMS, ERP systems
- Connect via the MCP server
- Or via your AI assistant
These are working hypotheses, not a finished roadmap. We are refining them with the operators in our beta.
Ray
Ray is our agent inside AMMP OS. Ask it for an overview and it builds the chart, then explains what it found on the way there.
Where our agents are heading
We are building a small team of specialised agents on top of the MCP server. Here is what they are being designed to do.
A coworker in Slack and Teams
A conversational agent that remembers your portfolio, learns from your feedback and pulls together answers from several data sources without you switching tools.
Visualisation on demand
Charts, dashboards and exports generated from a single question inside AMMP OS, instead of a request to a reporting queue.
Proactive QA and monitoring
Agents that run their own checks on data, metadata and dashboard integrity without being prompted, and flag what looks wrong before you notice it.
Escalation for hard problems
When a diagnosis is genuinely complex, the work is handed to a more capable reasoning model rather than answered with a guess.
Want early access?
We are onboarding a small group of asset managers and O&M providers. Tell us about your portfolio and we will show you the agents against your own data.
Get in touch