Now open source
twigex
AI Agents
Coming very soon

AI that acts on evidence,
not guesses.

Local AI agent framework with forced RBAC, full audit trails, and quality guardrails. Every answer cited back to its source. No prompt, document, or data ever leaves your infrastructure.

A Twigex AI Agents retrieval screen answering a question with a grounded answer and cited evidence

What makes it different.

  • Runs fully on your own infrastructure. No data leaves the building.
  • Forced role-based access control. Agents can only see what the requesting user is allowed to see.
  • Full audit trail of every agent action, query, and answer.
  • Quality guardrails. Only verified answers. If the knowledge graph cannot support a claim, the agent says so.

The legal and compliance angle.

One federal court has already ruled that privilege can be lost when a public AI tool's terms allow the provider to retain submitted data. Every prompt sent to a public AI tool is a potential disclosure.

Twigex AI Agents run entirely on your infrastructure. No prompt, document, or case fact is ever sent to a third-party model provider. The evidentiary record regulators increasingly expect is built in.

A preview of what's shipping

Every answer, backed by its evidence.

An AI Agents chat interface answering a net revenue question, showing the query plan, cited answer, and evidence panel with linked metrics, documents, entities, and relations
Ask a question, get a cited answer

Ask in plain language and the agent shows its work: which metric, entities, relations, and documents it drew from, with each claim in the answer linked back to its exact source. Nothing gets asserted without evidence you can open and check yourself.

Join the early access list.

AI Agents is in development alongside the Knowledge Graph. Sign up for priority access.

This opens your email client with the message pre-filled to wecare@twigex.com.