ITHZ MCP

- Different from generic RAG: the project owns the memory zone and the agent asks for bounded context packs.
- Different from chat history: decisions, risks, gates and reviewer notes are stored as project records instead of being buried in a conversation.
- The ledger layer lets a project allow, limit or block claims so agents can distinguish what may be said from what still needs review.
- Large-repo mode prunes generated codec/media/build outputs and keeps intake compact.
- Windows and POSIX host installers cover Codex, Claude Code, Cursor, Antigravity and generic JSON-RPC clients; macOS uses Python stdio until a macOS native binary is available.

A project record, not hidden chat memory

The MCP layer is designed around a simple boundary: code stays in Git, short startup rules stay in project.md, and long-lived agent work memory is stored in the local project.ithz archive. That archive can keep decisions, gates, risks, next steps and checkpoints close to the project instead of scattering them across chats or vendor-side memory.

Project Ledger v1 adds a more explicit control surface on top of that memory. A team can record durable decisions, scoped claims, blocked claims, reviewer notes and replication packs. In practice, this helps an agent know not only what the project remembers, but also what it is allowed to claim, what is still conditional and what must not be repeated publicly.

This does not turn ITHZ MCP into a production database, cloud sync product or Git replacement. Git remains the transport and history for source code. ITHZ MCP complements it with auditable work memory: bounded context packs, provenance, redaction checks and a clearer handoff trail for AI-assisted development.

For a team, the practical value is less context drift. A new agent run can start from project status, workflow profiles and a deterministic context pack before it reads broad source trees. Reviewers get a trail of why something was done, which gates applied, and which claims are safe or blocked.

The public package also keeps the safety boundary visible: raw prompts are not the default public memory surface, sensitive content can be redacted or blocked, and large repositories can avoid turning every task into a full-scan exercise.

MCP is about project-owned memory. The useful next step is to test the package and then understand how Native keeps the archive side inspectable.

Persistent memory, teams, and projects for AI agents. 76 MCP tools for storing, recalling, and sharing knowledge across sessions.

Persistent memory for AI assistants and coding agents across ChatGPT, Claude, Cursor, and other MCP-compatible tools.

Local Work Model for AI agents that learns from real outcomes.

Auditable, self-improving knowledge & memory for AI agents over MCP — citation-enforced answers and a replayable why-trace, self-hosted on Postgres.

One shared, versioned memory for your AI agents across Claude Code, Cursor, Kiro and Grok, with a source on every answer and full history.

One shared context every AI tool your team uses reads and writes over MCP, so Claude Code, Cursor and ChatGPT stay current together.

Open-source LLM knowledge base that turns team chat (Slack, Discord, Teams, Mattermost) into a typed knowledge graph and auto-generated wiki, exposed via a 28-tool MCP server.

Your company's brain, connected to Claude, ChatGPT, Gemini, Cursor, and VS Code. Turn your team's email history into shared memory that any AI assistant can query with natural language.

Operational memory for AI agents that persists across sessions and tools.