Every ingredient your AI can work with. Each one is useful on its own — but they're designed to work together, so skills can combine them into real workflows.
The people, organizations, and history that make up your world.
Everyone in your world — colleagues, clients, contacts, collaborators. Each person has a profile with how you know them, how important they are, when you last spoke, and any context your AI has captured over time. Import from LinkedIn, add manually, or let your AI build profiles from conversation.
These aren't contact cards. Each profile carries structured data: education history, work experience, certifications, publications, patents, languages, and skills — inferred automatically from resumes and LinkedIn. Every person tracks an importance level, a relationship type (contact, lead, prospect, champion, advisor, and more), outreach cadence settings, and contact details across multiple channels. Your AI can search, filter, and reason over all of it.
→ Steve, your product champion at SmallCo, just started as VP of Sales at BigCo. You know — because profiles link people to organizations with roles and dates, and your AI surfaces the change.
Companies, nonprofits, agencies, clients — any organization you're tracking. People link to orgs with roles (employee, advisor, board member, founder), so your AI can answer "who do I know at Acme?" or "which healthcare companies are in my network?"
Each org tracks pipeline stage (prospect, qualified, proposal, negotiation, closed), industry, website, and its own activity history. Orgs also link to each other — competitor, partner, parent company, subsidiary — so your AI understands the landscape, not just individual companies. When three of your contacts all work at the same place, your AI knows you have internal coverage there.
→ "Which prospects have I not talked to in two weeks?" Your AI checks pipeline stages, cross-references activity dates, and flags the ones going cold — without you building a dashboard.
Connections between people — not just between you and them, but between them and each other. "Alex introduced me to Jordan." "Sarah and Ben worked together at Stripe." These links turn a contact list into an actual network graph that your AI can traverse.
Relationships carry type (introduced-by, mentor, collaborator, peer, friend, alumni cohort, and more), confidence scoring, and notes. They're directional and timestamped. Your AI uses them to walk the graph — finding paths like "you know Alex, who introduced you to Jordan, who sits on the board at TargetCo." That's not a feature you'd build in a spreadsheet. It's what makes warm intros actually discoverable instead of something you stumble into.
→ Without relationships, you have a list of 500 contacts. With them, you have a network your AI can navigate — finding the three-hop path to anyone in your extended graph.
The history of what's happened between you and people in your network. Calls, emails, meetings, messages, notes — logged against a person or org so the context lives there permanently.
Each activity tracks type (call, email, meeting, demo, message, note), direction (inbound, outbound, internal), importance, channel, and scheduling. Your AI sets outreach cadence per contact — configurable days between touchpoints — and auto-detects when someone is going cold based on the gap since your last outbound activity. No manual tracking. The system knows when you're overdue.
→ Six months from now: "What's my history with Marcus?" Your AI pulls every call, email, and meeting — with dates, notes, and what was discussed — not just that you "talked sometime in Q1."
Conferences, meetups, dinners, job fairs — any gathering where people connect. Link attendees and organizations to events so your AI can prep you before and help you follow up after.
Events link to both people and organizations — so your AI can cross-reference the guest list with your existing network before you walk in. "Three attendees are second-degree connections — here's who could introduce you." Afterward, batch-create profiles for everyone you met, log activities, and set follow-up reminders in one conversation. Events become the origin story for relationships your AI tracks long after.
→ Six months post-conference, you need an intro to someone's company. Your AI already has the path — because it knows you met them at that event, who else was there, and who connects you.
How your AI reads, writes, tracks, and remembers across sessions.
Work items your AI can create, track, and update. Every "I should follow up on that" becomes a real task — not something you hope to remember. Organize into epics, flag blockers, add comments, and track status from backlog to done.
Tasks have real structure: epics break into tasks and subtasks. Each one flows through a kanban lifecycle — backlog, todo, in progress, review, done. Tasks can block other tasks, with reasons tracked. Threaded comments preserve context across sessions. And "asks" are a distinct primitive — cross-team requests with urgency levels (low, medium, high, urgent) and answer tracking, so "I need the logo from marketing by Friday" doesn't get lost in chat.
→ "What's still open from last week?" Your AI checks the board — and tells you two tasks are blocked, one ask is overdue, and the epic is 60% done.
Time-based nudges your AI can create, check, snooze, and complete. "Remind me to follow up with Sarah next Tuesday" becomes a real reminder linked to Sarah's profile — not a sticky note you'll lose.
Reminders link to profiles, organizations, or tasks — so the context travels with the nudge. Your AI doesn't just say "follow up with Sarah" — it knows which Sarah, what you last discussed, and what the reminder is about. Reminders surface automatically in your morning brief, support snooze and dismiss, and your AI marks them done when you act. They're the glue between "I should do this later" and actually doing it.
→ A reminder linked to Sarah's profile means your AI knows the who, the what, and the why — not just the when.
Markdown documents your AI can read and write — not just view. Use them for playbooks, briefs, research notes, meeting agendas, or anything you want to persist and build on over time. Your AI can also install skills — reusable workflows — directly into your workspace docs.
Docs are organized by folder — playbooks, research, templates, briefs — and your AI can create, read, update, and search them by meaning. This isn't file storage. Your AI actively uses docs mid-conversation: pulling up your intro email template when drafting outreach, updating a research brief after a discovery call, or writing a meeting summary you can reference next week. Semantic search finds relevant docs by concept, not filename.
→ Your "intro email playbook" gets pulled up automatically when a skill needs to draft an introduction — because your AI can read the doc and apply it, not just know it exists.
Your AI keeps a work diary — logging decisions, progress, and learnings as you go. At the end of a session, it writes a handoff: what happened, what's in progress, what to pick up next time. The next session starts oriented instead of blank.
The daily log captures typed entries — work completed, decisions made, things learned, blockers hit, plans for tomorrow. The handoff is structured: completed items, hot items that need attention, blockers, context updates, and tomorrow's focus. Your AI reads the previous handoff at the start of every session. For teams, handoffs sync between people — so when you pick up a project your collaborator was working on, your AI already knows what they did and what's next.
→ Monday's session picks up exactly where Friday left off. Your AI reads the handoff and knows what's done, what's stuck, and what to tackle first — no "where were we?" warm-up.
How your AI finds patterns, surfaces insights, and stays ahead.
Your relationships form a graph — not just who you know, but who they know, and how everyone connects. Your AI can traverse it to find warm introduction paths, identify your most connected contacts, and measure your reach at each degree of separation.
The graph engine finds warm intro paths to people and organizations, and can trace the shortest path between any two people up to six hops out. It scores super-connectors — people with the most bridging relationships across your network. It measures your reach at each degree of separation: how many people can you get to through one introduction versus two versus three? This is the difference between "I have contacts" and "I understand my network."
→ "Find me a path to the VP of Engineering at Stripe." Your AI walks the graph: you → Alex (college friend) → Jordan (Stripe eng manager) → target. Two hops, both warm.
Semantic search across your entire workspace — people, organizations, docs, logs, and notes. Your AI doesn't just match keywords; it understands meaning.
Search is embedding-based — it matches concepts, not strings. "Supply chain expert" finds someone whose profile says "logistics manager at a 3PL." It searches across every entity type: profiles, organizations, docs, daily logs, activities. Your AI uses it automatically when answering questions about your network or workspace, and skills use it to pull in relevant context before taking action.
→ "Who in my network understands enterprise sales cycles?" Found — even though no profile contains those exact words. Because meaning matters more than keywords.
Track where organizations sit in your relationship pipeline — prospect, qualified, proposal, negotiation, closed. Your AI can check pipeline health, surface deals that have gone quiet, and flag accounts missing a champion.
Pipeline tracks per-org stages with timestamps, so your AI knows how long each deal has been sitting. It detects stalled conversations — orgs stuck in the same stage with no recent activity. It flags orgs missing a champion — someone on the inside advocating for you. Combined with activities and people, pipeline gives your AI the full picture — not just where a deal is, but whether it's actually moving.
→ Your AI flags that Northfield has been "active" for six weeks with no activity — and your last touchpoint was a voicemail three weeks ago. CRM intelligence without the CRM.
Each ingredient is useful alone. Skills combine them into something qualitatively different.
You have a call with a prospect. You tell your AI "just spoke with Rachel at Northfield — she's interested, wants to see pricing next week." Your AI logs that as an activity on Rachel's profile and creates a reminder to send pricing. A week later, your morning brief surfaces it before you even ask. When you need a reference from someone similar, it checks your network graph and finds someone Rachel might already know.
You want to meet someone at a company you've been watching. Run /intro-path and your AI checks your contacts, finds people who work there, and maps out who knows who. It surfaces that your contact James went to school with their head of partnerships. It creates an outreach reminder and pulls up your "intro email playbook" from docs so you can draft the note. One command, five ingredients.
You just got back from a conference with a stack of business cards. Tell your AI who you met — it creates profiles, links them to the event, logs activities, and sets follow-up reminders for each one. Next week when you run /daily-brief, those follow-ups surface automatically. Six months later, you need an intro to one of their companies — your AI already has the path.
Your API token is what lets your AI reach into this workspace. Give it to Claude via MCP, or any other AI tool via the API. Once connected, your AI has read and write access to every ingredient — and the pantry grows richer every conversation.