Founding Member
Become a founding member by pre-ordering Terum for $20
-- OF 100 SPOTS LEFT
Here's what you'll unlock:
You're on the list!
Email added to waitlist — we'll email you with updates!
Join the waitlist
We'll notify you when early access opens.

One context layer. For you and your team.

Terum observes how your team actually works across every tool, then routes the right context to the right person or AI agent. No prompts to write, no agents to build, no docs to maintain.

You're in — we'll email you when early access opens. Contact Us
Built for Founding teams Dev squads Research labs Studios
TERUM
YOU
Home
Identity
Map
Chat
Memory1,247
Pulse
CONNECTIONS
Sources8
Apps5
RL
Ryan LiuPRO
Home YOU, TODAY
Dreaming
Good afternoon, Ryan.
Terum has observed you across 5 tools and 312 queries. Your identity is 67% known, and each session adds depth.
YOUR IDENTITY01
Strategic Builder
You think in systems, build with precision, and lead through clarity.
67%
strategywritingconcise
Updated todayView Identity →
WHAT'S STOREDLIVE
1,247
MEMORIES
Skills, preferences, topics, and patterns, continuously refined.
214 skills389 topics122 prefs
LiveExplore Map →
Claude 42%Cursor 24%ChatGPT 15%Notion 11%Gemini 8%
Live StreamLIVE
Refined writing pattern: prefers structured briefs2m ago
Claude session ended, 14 memories extracted5m ago
New topic detected: brand positioning12m ago
Preference confirmed: bullet points over paragraphs18m ago
Identity
Your archetype, traits, and behavioral patterns.
PRIMARY ARCHETYPE
Strategic Builder
You are the kind of mind that sees architecture where others see chaos. You build systems, lead through clarity, and bridge analysis with action.
CONFIDENCE 67%SESSIONS 312SOURCES 5
Analytical Thinking88%
Depth-focused74%
Detail-oriented71%
Solo Worker62%
Skills
Tools
Topics
Prefs
Good morning, Ryan
Your context is loaded. What are you working on?
Sonnet 4.5
RecentsProjectsStarred
Board meeting prep
Attendee context, agenda, and talking points for the 2pm sync.
Claude 4.5 Sonnet11 memories · 2h ago
Q1 investor update
Three-beat structure: wins, misses, ask. Declarative tone, no hedging.
Claude 4.5 Sonnet7 memories · 5h ago
API rate limiting
Token bucket with sliding window. 1000 req/min baseline, burst to 2000.
GPT-4o3 memories · Yesterday
Hiring pipeline review
Two senior roles open 90+ days. Switching to retained search.
Claude 4 Opus5 memories · Yesterday
Strategy
Go-to-market planning, competitive analysis, positioning work.
4 conversations18 memories
Product
Roadmap decisions, feature specs, integration planning.
3 conversations12 memories
Fundraise
Investor updates, pitch deck iterations, financial modeling.
2 conversations9 memories
Operations
Hiring pipeline, team structure, process improvements.
3 conversations8 memories
★ Board meeting prep
Attendee context, agenda, and talking points for the 2pm sync.
Claude 4.5 Sonnet11 memories · 2h ago
★ Q1 investor update
Three-beat structure: wins, misses, ask. Declarative tone, no hedging.
Claude 4.5 Sonnet7 memories · 5h ago
★ API rate limiting
Token bucket with sliding window. 1000 req/min baseline, burst to 2000.
GPT-4o3 memories · Yesterday
Memory
1,247 memories across skills, tools, topics, and preferences.
AllSkillsToolsTopicsPrefs
You write fastest in 90-minute blocks before noon, using structured briefs as launch pads.
writing3 days agoClaude
Strategy and AI Workflows overlap heavily — 71% shared context — your strongest knowledge cluster.
strategy5 days agoNotion
Prefers criticism first, fix second. Dislikes sugar-coated feedback or hedging language.
feedback1 week agoChatGPT
Cursor usage patterns show async/await preference, functional style over classes, and minimal comments.
cursor1 week agoCursor
Pulse
Your weekly rhythm and cognitive patterns.
47.2h
Total Hours
10:00am
Peak Window
6
Deep Work Blocks
72%
Focus Score
6
Artifacts Shipped
Deep workFragmentedMeetingIdle
9:00 – 10:321h 32mQ3 Campaign Strategyflow
10:45 – 11:2035mCode review, API authreview
11:30 – 12:1545mBrand positioning deckflow
13:00 – 13:4545mTeam sync, roadmapmeeting
Sources
8 connections active. Memory sync healthy.
CClaude
482 items · Synced 2m ago
CuCursor
318 items · Synced 5m ago
GChatGPT
156 items · Synced 12m ago
NNotion
89 items · Token expires in 3d
GeGemini
74 items · Synced 1h ago
GHGitHub
62 items · Synced 30m ago
LiLinkedIn
48 items · Synced 2h ago
GDGoogle Drive
18 items · Synced 4h ago
Apps
5 apps reading your context. 312 queries this week.
5
Connected Apps
312
Queries This Week
94%
Context Coverage
CClaude131 queries
2m ago
CuCursor75 queries
5m ago
GChatGPT47 queries
18m ago
NNotion34 queries
1h ago
GeGemini25 queries
3h ago
The Problem

Every AI tool on your team is individually smart, and collectively ignorant.

Your team already coordinates across Slack, Cursor, Linear, Notion, and four different AI providers. Every one of them starts from zero. Your people are the integration layer.

01

Individually smart, collectively ignorant

Every AI tool on your team starts from zero. The agent in Cursor doesn't know what was decided in Slack. The one drafting an email doesn't know the constraints from yesterday's Claude session.

02

Your team is the integration layer

Re-explaining project context in standups. Searching for decisions someone already made. Watching two people solve the same problem without knowing. Five hours per person per week, lost.

03

The wiki nobody updates

Documentation always lags behind reality. Decisions made in Slack last week never make it into Notion. The "how we do things" lives in the team's heads — and walks out the door with whoever leaves.

Why we're building this

We met at Harvard. In our separate medical imaging research labs — before we even knew each other — we hit the same problem. Teams tripping over each other's work. Hundreds of hours lost just keeping everyone's work in sync.

Then we started building a company together and hit it a third time. Even with just two people, we were losing hours a week to context friction. At some point we couldn't keep ignoring it.

Context flows from the bottom up.

— Teddy & Ryan
If you're a research lab, founding team, dev squad, or studio where every person is on the critical path — we built this for the version of you we were three years ago.
How it works

One pipeline, built from everyone's work up

Each teammate sets up once. Terum observes their work, combines it with everyone else's, and delivers context to the teammate or AI agent who needs it next. No instruction-writing, no agent-building, no docs.

10min Up and running in your browser in 10 minutes. Useful context by the end of your first day.
01

Observe

MCP, browser extension, and native OS integration. Runs in the background for each team member. Raw data stays on-device.

Connected Sources
G
Google DriveConnected
Li
LinkedInConnected
GH
GitHubConnected
Sl
SlackConnect
02

Structure

Raw activity becomes structured context: skills, preferences, workflows, decisions, solutions — per person and across the team.

Building Profile
Skills
Preferences
Workflows
Tone
Team context
Strategic thinker Data-driven Async-first
03

Deliver

Structured context delivered to any platform — and to the right teammate or AI agent when they need it. No manual prompting.

Context Delivery
LIVE
SYS Role: Head of Growth. Tone: concise. Team: Sarah's research (May 12), Marcus's API constraints.
YOU Help me write a proposal
AI Based on Sarah's findings last week...
You decide what's shared. Team leads see patterns, never individual activity. Every access is logged. Revoke any sharing in one click.
Day 3 after install
2:47 pm
Sarah opens Cursor to debug an API rate-limiting issue.
2:47 pm
Terum recognizes Marcus solved the same issue last Thursday.
2:48 pm
Marcus's approach and constraints surface in Sarah's IDE — before she spends an hour rediscovering it.
90 seconds instead of 90 minutes. Nobody had to write anything down.
Context Delivery

What delivery actually looks like

Same prompt, completely different output. You don't do anything differently — and team context flows in automatically.

For AI Agents

For you. For your team. For your AI.

When Marcus asks Claude for help, the agent inherits Marcus's context. When the team asks "what's our approach to X?", it inherits the team's. Scope is automatic — no one writes instructions, no one wires up data sources.

Workflow Automation

Terum spots what to automate. n8n runs it.

When Terum sees the same three-step sequence after every code review, it surfaces an n8n workflow and asks if you want to turn it on. Detection lives in Terum; execution lives in the automation tools your team already uses.

Detection, not configurationPatterns are surfaced from what your team already does — no triggers to wire up, no Zapier weekends.
Execution via n8nApproved workflows are handed off to n8n (or your existing automation platform). Terum doesn't compete with the plumbing.
Suggestion before actionEvery detected pattern lands as a proposal. Irreversible actions need a second confirmation. Every run is in the audit log.
Live Feed
Background actions
What you get

Four pieces that change how your team works

AI

Cross-person delivery

When Sarah figures something out, Marcus gets it in his Cursor session — before he asks. Context routes itself to the teammate or AI agent who needs it.

Team memory

Decisions, conventions, and solutions composed from each person's context. Survives turnover. Stays current. No one has to write it down.

Privacy-first architecture

Raw data is processed locally on each teammate's device. Only structured context syncs. Each person controls what reaches the team layer.

Admin controls + audit

Per-person sharing controls. Team-wide audit log. Every access logged. Revoke any integration in one click.

Landscape

Memory exists everywhere. No one captures how your team actually works.

Axis Dust Savant Pulse / Model Memory Terum
Target Enterprise (3,000+ orgs) Team / enterprise Individual Small team (2–50)
Agent creation Manual no-code builder API after extraction Single provider only Automatic from team behavior
Context capture Connected data sources Top-down extraction Chat history (one app) Passive observation across tools
Proactive delivery Search-pull Queryable ~ Personal briefing ✓ Right person, right tool
Team coordination Via agent fleet design Via shared operating context Native cross-person flows
Automation Agent-executed actions Decision-time serving Personal tasks Pattern-detected, n8n-executed
Setup effort Hours to days per agent Weeks of onboarding Zero Minutes (browser ext)
Dust orchestrates agents. Savant extracts knowledge. Terum just watches your team work.
Pricing

Free for individuals. Priced per teammate.

No credit card required. No trials that expire.

Free

$0forever

Your own context, across the AI tools you already use.

  • Browser extension
  • Up to 3 connected sources
  • Weekly context sync
  • Basic knowledge graph
Recommended

Pro

$18/ month

Live sync, native app, and personal AI agent context via MCP.

  • Everything in Free
  • Unlimited sources + real-time sync
  • Native desktop app
  • Personal AI agent context (MCP)
  • Proactive suggestions
  • Audit log

Studio

$39/ month

For solo founders, consultants, and researchers running heavy AI workflows.

  • Everything in Pro
  • Client / project compartments
  • Unlimited automations
  • Custom MCP server connections
  • Priority sync + processing
  • Beta features access
FAQ

Common questions

Terum observes your AI conversations, documents, code sessions, and tool usage. It extracts skills, preferences, topics, and workflow patterns, then structures them into a knowledge graph. Raw data is processed locally on each person's device. Only structured context syncs. On team plans, each person controls what they share to the team layer.

Capture Conversations and tool usage
Structure Knowledge graph extraction
Deliver Context to every person and tool

ChatGPT's memory only works inside ChatGPT, for one person. If you switch to Claude, Cursor, or Gemini, you start from zero again. Terum captures context from all your tools and coordinates it across your whole team and every AI tool they use, so every platform already knows how the team works.

Platform memory
Locked to one toolSimple key-value pairsResets across platformsYou manage it manually
vs
Terum
Works across every toolStructured knowledge graphSyncs automaticallyLearns and acts on patterns

Dust is an AI agent orchestration platform — powerful, but someone on your team has to build each agent, write its instructions, and connect its data sources. A 5-person startup doesn't have an AI operator.

Savant captures your team's documented and undocumented knowledge — but someone still has to identify what's worth capturing. The unwritten rules wait until someone notices them.

Terum watches how your team actually works — across every tool — and learns the patterns, decisions, and conventions on its own. When an agent needs the team's context, Terum hands over everything it has already learned.

No agent-building. No knowledge extraction. Nothing to write down.

No. Team leads see team-level analytics — knowledge flow patterns, coordination metrics — but never individual raw activity. You control what you share to the team layer. Every access is logged in your audit trail.

Never. Terum is funded entirely by user subscriptions. There is no advertising, no data brokering, and no third-party access to your context. Every query to your data is logged, and you can revoke any integration's access in one click.

Subscription-funded
No third-party access
Full audit log
One-click revoke

No. Terum is the context layer that makes every AI tool on your team smarter — regardless of provider. The browser extension works with any web-based AI tool today. A native desktop app is coming for Cursor, Figma, and VS Code. AI-native tools can connect directly via MCP. Your team's knowledge is never locked into one provider's memory.

Launch platforms
ChatGPTClaudeGeminiCopilot
Coming soon
CursorNotionFigmaVS CodeGitHubSlack