Free NGO Toolkit — get it now
// TECHNOLOGY

A controlled AI architecture built around your organisation.

See how organisational knowledge, AI reasoning and your existing tools connect — with a separate workspace and access boundary for each client.

// THE STACK

Five layers, each with a clear job.

Operational data and documents provide the context. Claude reasons over that context, while controlled connectors determine which systems it can access and what actions it can take.

Claude Projects

Layer 1 · Per-client workspace

An isolated workspace per organisation — its own instructions, knowledge and access boundary. No data bleed between clients.

Claude (Anthropic)

Layer 2 · AI orchestration layer

The reasoning engine. Understands context, retrieves knowledge, and decides which tools to call to complete a task.

MCP servers

Layer 3 · Tool-connection infrastructure

Custom Model Context Protocol servers expose your operational tools to Claude as callable actions — securely and per-client.

Qdrant

Layer 4 · Vector knowledge store

The vault. Documents are embedded and indexed for semantic retrieval, with nightly re-ingestion of staff corrections.

ImpactOS micro-apps

Layer 5 · Operational data layer

Lightweight apps that capture field and programme data at the source — the raw material the rest of the stack reasons over.

Common connector targets

Google WorkspaceKoBoToolboxZohoSalesforceMicrosoft 365
// DIRECT ANSWER

How does everyimpactAI connect knowledge, AI and operational tools?

everyimpactAI uses Claude for reasoning, Qdrant for searchable organisational knowledge and custom MCP connectors for controlled access to operational tools. Lightweight apps can capture programme data at the source, while each client receives a separate workspace and configuration boundary.