# René Zander

> René Zander builds context layers for AI agents on temporal knowledge graphs (Graphiti): sourced answers across chat, tickets, wiki and CRM, read with each person's own permissions, and actions authorized through explicit approval or scoped standing consent, on the client's own tenant. Knowledge graph consultant for DACH companies, based in Berlin-Brandenburg, working in German and English.

- Offer: knowledge graph consulting for context layers. A free 30-minute architecture call, a write-path review in a set number of hours, a phased build on the client's tenant billed by the hour, then operation.
- Where it runs: the client's own cloud tenant. Model endpoints, versions, deployment types and processing locations are verified for each engagement; self-hosted inference is an option.
- Stack: Graphiti on FalkorDB or Neo4j, MCP doors for assistants, approval cards in chat.
- Open source: Graphiti Local, a local-first temporal knowledge graph with six read-only MCP tools, a kg CLI and human-gated writes. An independent community project built on Graphiti, not affiliated with or endorsed by Zep.
- Background: 19 years of enterprise IT, former Solution Architect and Enterprise Architect at a Fortune 500.

## Offer

- [Knowledge graph consulting](https://renezander.com/knowledge-graph-consulting/): the context layer for AI agents on a temporal knowledge graph, and the way in.
- [Knowledge-Graph-Beratung](https://renezander.com/de/knowledge-graph-beratung/): the same offer in German.
- [AI agent control](https://renezander.com/agent-control/): how approvals, receipts and reversibility run in production.
- [Production audit](https://renezander.com/audit/): the write-path review, in a set number of hours.

## Free tools

- [AI Solution Shape](https://renezander.com/tools/ai-solution-shape/): describe a situation in a few sentences and see which kind of AI fits: automation, search, an AI-assisted workflow, a copilot or agents. A small model runs in the browser.
- [AI Decision Rights Map](https://renezander.com/tools/ai-decision-rights/): an approval matrix for AI agents. Per action: act alone, draft for a person to approve, recommend only, or never; with condition, named approver and whether the approval is checked again when the action runs. Flags six gaps and three go-live blockers. German: https://renezander.com/de/tools/ai-decision-rights/
- [AI Oversharing Check](https://renezander.com/tools/ai-oversharing-check/): an oversharing assessment before Copilot or any AI assistant goes live. Per source: who can open it (named people, one team, whole company, anyone with the link), what sensitive data sits in it, when access was last reviewed and who owns it. Flags five gaps (e.g. sensitive files shared by link, sensitive data open to the whole company), lists what to fix first, the safest first source for a pilot and three go-live blockers. Runs in the browser, EN and DE (/de/tools/ai-oversharing-check/).
- [All tools](https://renezander.com/tools/)

## Guides

- [Temporal knowledge graph](https://renezander.com/guides/temporal-knowledge-graph/): how facts with validity windows let AI agents answer from the current state, how Graphiti builds them, and when to use one.
- [Temporaler Wissensgraph](https://renezander.com/de/guides/temporaler-wissensgraph/): the same guide in German.
- [Graphiti vs Mem0](https://renezander.com/guides/graphiti-vs-mem0/): how the two memory layers store and update facts, what runs self-hosted, and which fits which agent.
- [Graphiti in production](https://renezander.com/guides/graphiti-in-production/): FalkorDB or Neo4j, Azure OpenAI or Ollama, embedding widths, version pins and ingestion cost.

## Projects

- [Graphiti Local](https://renezander.com/projects/graphiti-local/): local-first temporal knowledge graph with read-only MCP tools and human-gated writes. Source code under the Apache-2.0 license: https://github.com/renezander030/graphiti-local

## Proof

- [Case study: context layer for AI agents](https://renezander.com/case-studies/operational-context-layer-governed-actions/): an operational context layer over chat, tickets, wiki and CRM on a company's own EU tenant.

## Writing

- [Two kinds of agent memory](https://renezander.com/blog/agent-memory-okf-vs-codebase-knowledge-graph/): authored knowledge versus knowledge derived from code.
- [Agent memory approval gates](https://renezander.com/blog/agent-memory-approval-gates/): which memory writes never need a human, and which always do.

## Contact

- Email: rene@renezander.com
- Book a call: https://www.cal.eu/reneza/30min
- LinkedIn: https://www.linkedin.com/in/reneza
- GitHub: https://github.com/renezander030