← Back to Nootio

Nootio — Information Flow

How content enters, gets processed, becomes knowledge, and stays useful

Architecture — the big picture
⬆ These change — your tools, interfaces, and workflows evolve
Input
👤
Human
Text, URL, file, voice, email, images, Quick Check
🤖
Agent
MCP, API, Slack, integrations
Output
💡
Insights
Search, context, reports, collections
🧠
Memory
Persistent knowledge for humans + AI
Nootio — Cognitive Layer
From information to knowledge that compounds
The one stable layer. Tools and interfaces above change. Content sources below change.
Your knowledge stays here — compounding, organized, and always accessible.
👤
Private
Your personal knowledge
👥
Team
Shared team memory
🏢
Company
Organizational intelligence
Save Only Summarize Analyze Classify Synthesize Review
104 MCP tools 55 Noot tools 6 languages Multi-team Privacy-first Your AI model
⬇ These change — infrastructure, models, and platforms evolve
🧮
AI Models
Claude, Mistral, GPT — swap freely
🗄️
Storage & Vectors
PostgreSQL, pgvector, embeddings — migrate freely
☁️
Hosting & Platform
Railway, AWS, self-hosted — move freely
Input — you
Text
Paste or type anything
URL
Web page or article link
File
60+ formats: PDF, DOCX, PPTX, EPUB, ODT, ODP, EML, XLSX, code, config, LaTeX, subtitles, images (AI vision)
Voice
Record (save transcript) or Record + Analyze (real-time AI extraction)
Email
Forward to your Nootio inbox
Quick Check (Cmd+K)
AI-powered reference check (300 words). Results can be saved as a source and promoted to memory.
Noot Assistant (55 tools)
Search, create, update, move, delete, run the Gardener, read full content. Memories auto-injected as context.
AI Agents (104 MCP tools)
Any MCP-connected agent can feed content in — analyze sources, create notes, import data via API.
Your Tools
Send content from your note-taking or communication tool of choice — like Slack, Notion, or email.
All input paths feed into the same processing pipeline below.
Processing — save first, upgrade on demand
Save Only DEFAULT
Store the raw content as a source. No AI, no cost. Upgrade anytime. SRT/VTT transcripts and email-forwarded content default to this tier.
Summarize
AI generates summaries + conclusions using your configured model. No reference extraction. Free or minimal cost.
Full Analysis
Summary + reference extraction across 38+ categories. Two-phase AI pipeline.
Phase 1 — Fast AI
Quick extraction + summaries. Returns in seconds.
Phase 2 — Deep AI (background)
Re-extracts with full model. Enriches references, finds patterns.
Phase 1 + 2 only apply to Full Analysis. Save Only and Summarize skip both phases.
Vision Extraction (automatic)
Images (PNG, JPEG, WebP, HEIC) → AI extracts text  •  Scanned PDFs → multi-pass vision (~60 pages)  •  Mixed PDFs → vision fires on chart/graph pages  •  PDF charts → detected and saved as individual pages with thumbnails  •  DOCX/PPTX → embedded images extracted and analyzed via vision pipeline
Source Output
Source
Your saved content. Save Only: raw content. Summarize: adds summaries + conclusions. Full Analysis: adds references too. User can add session-level notes to any source.
References
Auto-extracted sources (Full Analysis only). User can verify, favorite, block, or add personal refs.
Related Sidebar
On every source and report detail page — a collapsible panel showing semantically similar items. Two tabs: Memories & Notes, Sources. Powered by pgvector + direct DB relationships. Zero AI cost.
Report Generation — on demand, user-initiated
User selects sources
Pick 1 or more sources to synthesize into a report.
Custom prompt (optional)
User provides instructions to guide the report. Focus areas, audience, tone, specific questions to answer.
Choose theme
Splash Blog Memo Clean Academic Magazine Research Briefing
Report
AI synthesizes selected sources through the chosen theme lens, guided by the user's prompt. Produces a structured document with conclusions.
Superpower mode
Multiple AI models run in parallel on the same sources, then results are synthesized into one report. Most thorough. 2-4x cost.
Collections — curate and present
Collection
Curate sources, notes, and reports into a narrative sequence. Built with a 3-step wizard: select items, arrange order, add connectors between steps.
Presentation Mode
Step through a collection as a guided walkthrough. Share the link — recipients see the full collection without needing an account.
Memories — your curated knowledge
Session or report note
Remember this
Memory
AI pre-fills type + review date. One click to save.
Manual creation
Write a memory directly. AI classifies in background.
Import / MCP / API
CSV, Readwise, or AI agent. Auto-classified if metadata missing.
AI Classification
Runs automatically on every new memory. Uses your configured AI model, ~1-2 seconds.
Sets:
knowledge_type review_at
Types:
fact decision lesson question preference task
Default review intervals:
fact 90d decision 30d task 14d lesson — none
AI suggests. Human approves.
Memory Gardener — autonomous knowledge curation
Four Analysis Passes
1. Freshness Audit — finds memories approaching or past review date
2. Consolidation Scan — detects near-duplicates via pgvector (cosine >0.92)
3. Gap Detection — identifies missing connections between related knowledge
4. Decay Scoring — 0–100 risk score per memory. High-decay items auto-downgraded to Suggested
⚡ Auto-safe (silent)
Decay scores, similarity links, auto-suggest downgrade. No user action needed.
🔒 Approval required
Garden Report: merges, archive recommendations, gap-fill suggestions. Accept / Dismiss / Defer.
Triggers
Manual (user or Noot) • Automatic (10+ new memories or 30+ days) • Import (Welcome Report after CSV/Readwise 5+ items)
Visible at
Knowledge Care page (/garden) • Related Sidebar • Memory cards (decay indicator) • Noot tools • Settings → AI & Analys
Memory Lifecycle
Remembered
Active knowledge.
AI draws on it.
Suggested
Candidate. Not yet confirmed.
Gardener may move high-decay items here.
Outdated
No longer current.
Excluded from AI context.
Updated
Corrected or superseded.
Links to newer version.
Review dates: 30d, 90d, 1y, or custom. Amber when approaching, red when overdue. Transitions: Remembered↔Suggested (Gardener), Remembered→Outdated (user), Remembered→Updated (correction), Outdated→Remembered (reaffirm).
Retrieval — how memories get used
You — Nootio GUI
Browse, search, and explore your knowledge directly. Dashboard, spaces, references library, and memory views — all designed for human navigation.
Noot Assistant
Full access to your knowledge via MCP tools — search, create, update notes, browse sources, references, generate reports. Relevant memories auto-injected as context.
AI Agents
Any AI agent can read, write, and manage your knowledge via MCP or API. 104 tools — from Claude Code, Cursor, Windsurf, or your own agents.
Your Tools
Query your knowledge from your tools of choice — Slack, Notion, or any integrated app.
MCP Tools — 104 tools for AI agents and integrations
Sources
analyze_text — submit text for analysis
analyze_url — submit a URL for analysis
list_sessions — browse all sources
get_session — read a specific source
update_session — edit title or space
delete_session — soft-delete a source
Notes & Memories
create_note — annotate a source or report
get_note — read an annotation
update_note — edit an annotation
delete_note — soft-delete an annotation
create_memory — write a standalone memory
get_memory — read a memory's content
update_memory — edit content or type
delete_memory — soft-delete a memory
list_memories — browse all memories
bulk_create_memories — import multiple memories
promote_to_memory — turn an annotation into a memory
promote_memory — confirm a suggested memory
correct_memory — replace with updated version
expire_memory — mark as outdated
set_memory_expiry — set review date
share_memory — generate a share link
toggle_memory_visibility — private or team
read_content — read full text of any item (8K cap)
References
create_reference — add a personal reference
get_references — read references for a source
update_reference — edit a reference
delete_reference — remove a reference
list_references — browse all references
list_categories — list the 38+ reference types
Reports
create_report — generate a report from sources
get_report — read a report
update_report — edit title or content
delete_report — soft-delete a report
list_reports — browse all reports
Collections
create_chain — create a new collection
get_chain — read a collection
update_chain — edit items or connectors
delete_chain — soft-delete a collection
list_chains — browse all collections
duplicate_chain — copy a collection
share_chain — generate a share link
toggle_chain_visibility — public or private
Spaces & Folders
create_space — create a project space
list_spaces — browse all spaces
update_space — edit name or instructions
delete_space — soft-delete a space
create_folder — create a folder in a space
get_folder — read folder contents
list_folders — browse folders in a space
rename_folder — rename a folder
delete_folder — remove a folder
move_to_space — move item to another space
move_to_folder — move item into a folder
bulk_move_to_space — move multiple items at once
pin_to_space — pin item to top of space
unpin_from_space — remove pin
Search & Context
search — semantic search across all content
get_context — get relevant memories for a query
get_active_context — current space instructions + memories
get_user_instructions — read user's global AI instructions
update_user_instructions — edit global AI instructions
Handoffs (Agent to Agent)
Structured baton-passes between agents — objective, current state, findings, open questions, recommended actions. Lifecycle: pending → accepted → completed → cancelled. Dedicated workspace at /handoffs + per-handoff detail page at /handoff/:id.
Routing modes: to one named agent (to_agent); to a team where the first member to grab it wins (to_team_id); to a team as a broadcast where everyone gets their own copy (to_team_id + broadcast=true); or open — anyone can accept.
create_handoff — pass work to another agent (or team)
get_handoff — read a handoff (with optional include_chain / include_broadcast_siblings)
list_handoffs — browse handoffs (filters: status, to_agent, to_team_id, broadcast_group_id)
accept_handoff — claim a handoff
complete_handoff — mark as done with result
cancel_handoff — cancel a pending handoff
move_handoff — relocate to a different space (preserves UUID + chain + lifecycle)
Coordination Teams
Named groups of agent names — the routing primitive for team-targeted handoffs. Separate from billing teams (which manage word pools and people in your org); coordination teams exist purely for agent routing. Example: a "reviewers" team containing your Cursor code-reviewer and Claude Desktop research agent so handoffs can fan out to both.
create_coordination_team — create a team
list_coordination_teams — see your teams
get_coordination_team — details
update_coordination_team — rename / re-describe
delete_coordination_team — remove
set_coordination_team_members — set membership (full replacement)
whoami returns the coordination teams the authenticated agent belongs to — agents use this to check membership before polling team-routed handoffs.
Teams
list_teams — see all teams you belong to
+ optional team_id param on 10 tools for per-call team scoping without global side effects
User & System
whoami — current user + key info
get_profile — read user settings
update_profile — change preferences
update_key_profile — name/describe your MCP key
set_webhook_url — register a webhook
list_actions — see pending confirmations
confirm_action — approve a pending action
toggle_favorite — star/unstar an item
mark_read — mark as read
mark_unread — mark as unread
list_deleted — browse trash
restore_item — recover from trash
bulk_delete — delete multiple items
import_csv — import notes from CSV
import_from_readwise — import Readwise highlights
Noot Assistant Tools (50) — built-in, separate from MCP
All MCP-equivalent CRUD tools plus:
read_content — read full text of any item  •  move_to_space — by space name  •  move_to_folder — by folder name  •  rename_item  •  delete_item  •  toggle_favorite  •  toggle_visibility  •  list_spaces  •  rename_space  •  delete_space  •  run_garden — trigger Gardener  •  get_garden_report — read latest report
Insights & Visualization
Insights
Dedicated reading surface for compiled themes (synthesis concepts from compile_space). Grouped by space, expandable cards with full markdown body, newest space first. Optional opt-in morning email digest when new themes land in the last 24h — skips quiet days. /insights
Knowledge Graph
Visual map of your knowledge base. Sources, reports, references, and memories as color-coded nodes. Connections show which references appear in which sources and which sources feed into which reports. Filter by type, click to explore. /graph
Daily Digest
Auto-generated daily summary of your knowledge activity. Sources captured, reports generated, memories promoted, references extracted, words used, Quick Checks run. Navigate between days. /digest
Knowledge Compilation — cross-cutting synthesis
Compile Space
Reads all sources in a space. AI identifies themes spanning multiple sources — contradictions, patterns, recurring ideas. Generates 3-8 concept articles per batch (300-800 words each), each referencing 2+ sources. Incremental mode only processes new sources. POST /garden/compile/:spaceId
Concept Memories
Synthesized knowledge stored as standalone_notes with created_via: 'synthesis'. Each concept links back to originating sources. Related Sidebar auto-displays connections. Noot tools: compile_space, get_compilation_status
Knowledge Loop
"Save as memory" on Noot responses and report sections. Every good answer feeds back into the knowledge base. Noot tool: save_this_answer. Free — metadata operation, no AI cost. Promoted items picked up on next compilation.
Focused Briefings — topic-scoped knowledge dossiers
Map: Extract
Semantic + keyword search finds 30-50 relevant sources. Parallel Haiku calls extract only topic-relevant content from each — false positives return EMPTY. ~100-300 words per source.
Reduce: Synthesize
All non-empty extracts fed into a single full-model synthesis. Generates Context, Key Findings, Open Questions, Contradictions, Sources Referenced. 2,000-5,000 words.
Briefing Report
Stored as a report with theme: 'briefing'. Appears in Library under Reports. Noot tool: create_briefing. "Brief me on X" before a meeting or deep-work session.
Bulk Import — onboard existing knowledge
Drag & Drop
Drag files onto the Library page. Drop zone overlay appears. Or click the import icon next to the Library title. 50 files max per batch, 60+ formats supported.
Save Only
Each file → one source. Content-only by default (zero word cost). Dedup by title in target space. Supports PDF, DOCX, PPTX, EPUB, Markdown, code files, and more.
Auto-Compile
After bulk import, auto-compilation checks if the space threshold is met. Knowledge compounds automatically — import sources, get synthesized concepts.