Tools we use first.
Internal tools that became products. Each one solves something we needed before we offered it to anyone else. Built provider-agnostic: Claude, OpenAI, Ollama, your own local model. Pick the engine that fits the engagement.
What is a virtual C-Suite for operators?
Seven AI executives that work as a boardroom or one-on-one. Ask a strategic question and the executives respond from their domains, then the Executive Hub synthesizes a recommendation. Or go direct with the one you need.
Built for MSPs, small teams, and founders running things. The executives carry context across conversations, files, and projects, backed by RhizomeRAG. The Boardroom keeps the receipts so the advice stays grounded in your business, not generic best practice.
Episodic memory for AI agents.
A networked memory layer built for AI agents that need to remember entities, relationships, and episodes across sessions. Query it, traverse it, or ground LLM responses against structured context instead of a flat vector blob.
RhizomeRAG is the substrate behind Hextant. We open-source it because the memory problem is bigger than any one product, and the agent community moves faster when the bones are shared.
The Syncro PSA, scriptable.
A typed TypeScript client, a command-line interface, and an MCP server for the Syncro MSP REST API. Drive tickets, customers, assets, invoices, and RMM alerts from a shell or from any AI agent that speaks MCP.
Built because the Syncro UI is the workflow, and the workflow slows down when you click through it. The SDK is what our own operators use to make Syncro fit how we actually run an MSP.
The engine behind Hextant.
A multi-agent orchestration engine. It dispatches parallel agent sessions, runs interactive multi-turn conversations with the Hextant executives, and chains workflows across Claude Code, Anthropic, OpenAI, and Ollama. One governance substrate, three modes (dispatch, sessions, workflows).
The infrastructure under every Hexaxia AI tool that talks to more than one model. Stagnation detection, cost enforcement per dispatch, and an audit trail per session live at the engine layer so the agents above stay focused on the work.
Documentation that earns a second life.
Doxia is a documentation cleanup pipeline with a useful second use. Point it at the knowledge base you actually have. AI-assisted normalization turns the sprawl into a clean, structured corpus. The same output that makes the wiki readable is also the substrate that makes the next AI step viable.
Forward flow cleans the org. Reverse flow ships a corpus that is primed for retrieval, ingestion, or training. The months of data prep that usually kill private-LLM and on-prem AI projects start here, with the docs you already have.
Doxia is in active development. The pipeline already runs against real client knowledge bases; we are hardening the security gates and the connector set before a wider release. It is proprietary, and built to run on-prem where the documents cannot leave the building.
What is being prototyped next.
Three threads in active concept work, all sharing the same substrate. None of them are pitched as products yet. We are working through them with the clients who need them most and will pull them into the lineup when the shape is real.
On-prem AI
LLMs that run inside the walls.
For clients who cannot ship data to the public cloud. The full Hexaxia AI stack, deployed against on-prem inference, with the same governance posture.
Private models
Trained on what is yours.
Fine-tuned models against organizational corpora. Not a chatbot with your logo on it. A model that speaks your domain because it was trained on it.
Training pipelines
Doxia output, repurposed.
The same clean corpus that comes out of Doxia is the input to fine-tuning. The piece that usually takes months becomes the part you already did to get a wiki you can read.
Products are what we ship. Engagements are how we deliver them. Practice areas are what gets built inside the engagements.