Where a governed shared memory earns its place

The situation

Consultants move between clients and use AI assistants all day. The firm needs the assistants to be useful on each engagement without any chance that one client's material surfaces in another client's work.

How cMCP helps Available now

• Client walls: a client's memory is returned only while working on that client or a project linked to it
• Audience labels keep restricted material to named people, and external_ok marks what is safe for outside replies
• Client-scoped memory inherited by everyone on the engagement, so a new consultant starts informed

The situation

A multi-year programme loses people to other projects and to retirement. Decisions, supplier quirks and lessons learned leave with them, and newcomers spend weeks asking questions that were answered years ago.

How cMCP helps Available now

• Insights confirmed by at least 3 colleagues become team knowledge, then project knowledge with a lead's approval
• New members inherit project context immediately through the context pack
• Versioning keeps the history of how a decision changed

The situation

In regulated work, "the assistant suggested it" is not an answer. The team needs to show which information was available when an AI-assisted decision was made.

How cMCP helps Available now

• The audit log records every allow and deny with who, where, which app and which model
• "What the AI knew" shows the memory versions behind each answer
• Governed writes mean shared memory changes only through a role, a grant or an approval

Roadmap Hash-chained, exportable audit is on the roadmap. cMCP does not claim any regulatory compliance.

The situation

Different teams prefer different assistants. Each one learns in isolation, and switching tools means starting from zero.

How cMCP helps Available now

• One memory behind open standards (MCP and REST), so any AI app that speaks MCP can connect
• Available now with Claude Code and Claude Desktop
• Memory is yours and outlives any single model or vendor choice

In testing Connecting ChatGPT, Gemini, Copilot, Cursor, Codex, Gemini CLI, Cline and local models is built and in testing. Anything that does not pass testing will be removed.

The situation

Several agents and digital workers contribute to the same project. Without shared state they duplicate effort, or worse, reach conflicting conclusions.

How cMCP helps Available now

• Bots and digital workers have their own identity and a human sponsor, and can never approve
• Agents read the project's shared memory through the context pack and propose changes through the approval flow
• Kill switches suspend a misbehaving bot immediately; an agent never has more rights than its user