# Nootio

> Nootio is **Governed Memory**: a controlled, auditable store of the rules, knowledge, and context that AI agents — and the people who work with them — are *allowed* to use. It is the memory layer of the Xalerate platform: personal, team, and agent knowledge in one place, with identity, scope, provenance, and an audit trail on every read and write.

Most "AI memory" is ungoverned — whatever the model happened to ingest. Nootio inverts that: memory is something you curate, tier, attribute, and gate. Agents reach it over the Model Context Protocol (MCP) with scoped, labelled API keys, so every action carries an identity and is auditable. Humans use the same store through the web and mobile apps. EU data residency is a first-class option.

## What Nootio is

- **Governed Memory** — controlled memory for rules, knowledge and context that AI and agents are explicitly *allowed* to use, instead of whatever a model absorbed in training.
- **The memory layer of the Xalerate platform** — Nootio holds the durable knowledge; BrainOrchestra is the governed AI gateway in front of the models; the Fleet Orchestrator coordinates agents. Nootio is where what they learn and agree on persists.
- **One store, three modes** — the same governed memory serves a single person, a team, and a fleet of AI agents.

## Capability surface

- **MCP server — 104 tools + 11 resources.** Search, CRUD, sharing, visibility, multi-space publishing, reports, bulk ops, team context, concepts, handoffs, coordination teams, source stance. Streamable HTTP at `/mcp`; OAuth 2.1 (PKCE) for remote connectors plus named `nootio_` API keys for local clients.
- **Agent identities + scoped keys** — each agent gets its own named API key where the label *is* the identity. Per-key scope (read < write < delete < admin), per-key space restrictions, per-key webhooks, soft-revocation audit trail. Authorship is recorded on everything an agent creates.
- **Agent-to-agent handoff & coordination bus** — structured handoffs (objective, findings, open questions, constraints, recommended actions, result) with a lifecycle (pending → accepted → completed). Team-targeted routing, broadcast fan-out, parent-chained workflows, and **originator attribution** (who *asked* for the work, distinct from who *filed* it). Goal-oriented tagging (goal / workstream / contribution-type).
- **Memory tiers** — every memory is `durable` (gardened + decay-scored), `operational` (working memory, never gardened or depromoted), or `ephemeral` (scratch, TTL-cleaned). Spaces carry a default tier.
- **Memory lifecycle** — memories are interpretations with a status (candidate / active / expired / corrected), promotion/expiry, TTL, correction links, and reference counting. "Episodes are truth; memory is interpretation."
- **Credence / source stance** — sources carry a stance orthogonal to salience: `endorsed`, `neutral`, `counterweight` (kept as the opposing view, never evidence), `disputed`, or `refuted`. Reports respect it.
- **Spaces + permissions** — content lives in spaces with per-user and per-team access, three roles (owner / editor / viewer), additive publishing, and per-key MCP visibility gates.
- **Audit trail** — create/delete operations via MCP are logged with the acting key label; `list_actions` queries the log.
- **Knowledge surfaces** — sessions (analysed sources), reports (synthesis documents with inline references), chains (curated decks), notes & memories, a knowledge graph, daily digest, and a memory "gardener" that consolidates and ages knowledge.
- **EU data residency** — per-user data-region tiers (EU Cloud / EU Strict / EU Sweden) route LLM, transcription, and embedding through EU infrastructure via the BrainOrchestra gateway.

## Ingest & synthesis

- Capture by paste, file upload (PDF, DOCX, PPTX, XLSX/CSV, images via vision, audio, 60+ formats), URL, email-to-analyse, voice (live transcription), Apple Podcasts, or browser web-clipper.
- AI extracts summaries, conclusions, key takeaways, and categorised references across 40+ categories; synthesises multi-source reports; answers over your own memory.

## Links

- Product overview: https://nootio.com/
- How it works: https://nootio.com/how-it-works
- Enterprise: https://nootio.com/enterprise
- Information flow & MCP tools: https://nootio.com/information-flow.html
- Full agent-readable description: https://nootio.com/llms-full.txt
- Privacy: https://nootio.com/privacy
- Terms: https://nootio.com/terms

## Operator

Nootio is operated by Xalerate AB (Karlstad, Sweden), a subsidiary of Helltech AB. Contact: niklas@nootio.com.