Awareness Use Cases: From Enterprise Memory to Personal IDE Sync

Awareness is a memory layer, so the question is never "what does it do" but "what do you want your AI to remember." These use cases are the ones we see most often — each one is a real deployment pattern, not a slide.

1. Enterprise: Turn legacy databases into an AI-readable brain

The problem. A mid-size enterprise has decades of data locked in a legacy ERP (SAP, Oracle EBS), a CRM (Salesforce or on-prem Dynamics), and office documents (Word, Excel, contracts). None of it is AI-readable: a new AI assistant has no way to answer "what did we quote Acme Corp in Q3?" or "which suppliers are on net-90 terms?" The data is there — it is just not connected.

The Awareness pattern. Instead of replacing the ERP/CRM, Awareness becomes the central memory brain that sits on top:

1. Migrate & vectorize — export the legacy databases and office files, then index them: every record becomes a searchable chunk, and every entity (customer, supplier, product, contract) becomes a node in the knowledge graph with its relationships.
2. One graph, one query surface — ERP records, CRM accounts, and document knowledge now live in the same graph. An AI agent can ask a single question across all three systems.
3. Let the AI read and write — through MCP or the SDK, agents retrieve context before answering and write back structured facts (a decision, a resolved issue) so the memory compounds.
4. Deploy on-prem — full local deployment for data residency (China PIPL, GDPR) and confidentiality; no third-party cloud involved.

Outcome. The company gets an AI central memory that understands internal information precisely: "Who are our top 10 customers by margin, and what did we promise each of them last quarter?" is answerable in seconds by an agent that cites the exact ERP + CRM + contract records. On-premises means the board is comfortable; connected means the AI is actually useful.

Example industries: manufacturing (inventory + orders + supplier contracts), logistics (shipments + SLAs), government (regulatory records), healthcare (patient records — with proper access control).

2. Individual developer: one memory across every IDE

The problem. A developer uses Cursor for daily coding, Claude Code for big refactors, Windsurf for quick experiments, and sometimes ChatGPT for design questions. Every tool is smart — but none remembers what the others did. Every morning starts with re-explaining the project.

The Awareness pattern. Because Awareness exposes memory through the standard MCP protocol, every tool connects to the same memory:

• Claude Code and Cursor read and write the same knowledge base through MCP
• Workflow rules are auto-injected so each agent knows when to recall and when to record
• Decisions, pitfalls, and architecture notes made in one tool are available in every other tool

Outcome. Multi-IDE collaboration without manual syncing: refactor a module in Claude Code on Monday, open the same project in Cursor on Tuesday, and the agent already knows the refactor decisions, the file layout, and the pitfalls you hit. One brain, many tools.

3. Law firm: one shared memory across Harvey and Microsoft Copilot

The problem. A law firm runs Harvey (the legal AI assistant built on GPT-4-class models) for legal research and drafting, and Microsoft Copilot for Office work (Word, Outlook, Teams). Associates switch between them constantly, and neither knows what the other was told. Context — client instructions, matter background, preferred jurisdiction — gets re-explained or lost.

The Awareness pattern. Awareness sits between them as a shared matter memory:

1. Matter workspace — each client matter gets its own memory space with access control per team role.
2. Harvey integration — research conclusions and drafting decisions are written back to the matter memory through the SDK or API.
3. Copilot / Office integration — Word documents and email summaries flow into the same matter memory; Copilot can pull the firm's prior work product.
4. Permission boundaries — different matters, different partners, strict confidentiality isolation (critical for privileged communications).

Outcome. Two specialized AI systems stop being two silos and start being one firm brain. A partner opens a matter in Word, Copilot drafts from the firm's actual prior work; an associate asks Harvey for research and the query is grounded in the same matter context. Work product compounds instead of vanishing into individual chat histories.

4. More patterns we see

• Support teams: tickets + knowledge base + customer history → agents that resolve issues citing past resolutions.
• Sales: CRM + call transcripts + contracts → "prepare the Q3 renewal pitch" with full account context.
• Recruiting: applicant tracking + interview notes → consistent candidate evaluation across a hiring team.
• Research teams: papers + lab notes + decisions → a shared lab memory that compounds across projects.

How to start

Every use case follows the same three steps:

1. Connect — install the MCP server (npx @awareness.market/setup) or use the SDK, pointed at your memory space.
2. Index — import documents or databases; Awareness chunks, embeds, and builds the knowledge graph.
3. Use — your AI tools (Claude Code, Cursor, ChatGPT, Harvey, Copilot, custom agents) now read and write one shared memory.

Self-hosted deployment for enterprises: see the Enterprise Deployment Guide.