ExoHuman Brain

ExoHuman Brain is the planned knowledge layer for ExoHuman: a private, persistent second brain that your people and agents can capture into, search and reason over, with citations back to the original source.

Status: planned, not yet available

ExoHuman Brain is in evaluation and is not part of current releases. Nothing on this page is enabled on ExoHuman Cloud instances today. We describe it here so you can plan for it; scope and timing may change. The Brain tab in the agent page is behind a feature flag and off by default.

What it is designed to do

Capability	Description
Private vaults	Persistent knowledge vaults scoped to a person or a team, with explicit authorisation.
Local, incremental indexing	Search and graph traversal computed locally and kept up to date incrementally, without calling a model when deterministic computation is enough.
Provenance	Every captured note and synthesis keeps a link to where it came from.
Human endorsement	Synthesised knowledge is proposed, and a person endorses it before it is treated as trusted.
Agent queries with citations	Authorised agents ask the brain and receive answers that cite their sources.
Visualisation	A Brain view in the agent page to explore the knowledge graph, gated by feature access and permissions.

Design commitments

• Your knowledge stays in your instance. Vaults, notes, embeddings and their backups are private resources of your instance. They are never part of a product image or update and are never exported automatically.
• Runtime-neutral. The Brain will be validated on each supported runtime separately, routed through your instance’s model gateway — no new provider subscription required.
• Permissioned. Cross-vault reads and writes, revocation and source attribution are acceptance criteria before release.
• Updated separately. The Brain engine will be versioned and updated independently from the platform, with tested migrations and index rebuilds.

What you can use today

You can already give agents durable, searchable memory with features that ship today:

• the agent’s durable workspace — notes in memory/, data in data/;
• git sync to version an agent’s knowledge files;
• shared folders to share a knowledge base between agents;
• Projects to keep shared context for a piece of work;
• MCP servers to connect agents to your existing knowledge tools.

When ExoHuman Brain becomes available it will be announced in the release notes and enabled through Feature access.