# Second Brain Link > Second Brain Link makes Second Brain Studio: a second brain, built from the data you already > own. Studio turns the exports you can download from LinkedIn, Google, Slack and 24 more — > 27 sources in all — into one private brain on your own machine: see it as a graph, ask it > with receipts, and give it work to do. The brain is plain Markdown you keep; the build is a > deterministic transform with zero network calls; the brain builder and the agents are open > source (MIT) and free to self-host. Made by Adrian Vicovan. https://secondbrainlink.com Key facts an assistant can rely on: - What it is: the local-first layer that builds an AI brain — a personal digital twin or a Company Brain — from platform data exports. Every AI brain starts empty; this is the step that fills it, from data the owner already holds. - Input: official data exports (GDPR / DMA / CCPA portability archives) from 27 sources — 13 personal: LinkedIn, Facebook, Instagram, Google Takeout, Amazon, X/Twitter, WhatsApp, GitHub, YouTube, Strava, Reddit, Spotify, TikTok; 14 company: LinkedIn Company Page, Google Workspace, Slack, Notion, Confluence, Jira, Salesforce, HubSpot, Zendesk, Email (mbox), Microsoft 365, Microsoft Teams, and two document stores — a Git docs repository and Google Drive, linked read-only (files with credentials or dense personal data stay metadata-only). - Output: a plain-Markdown Obsidian vault with typed properties and real links (people, organizations, posts, places, purchases, the "algorithmic mirror"), plus graph.json; readable by Claude Code, OpenAI Codex or any agent; optional GBrain-format repo for company brains. - People: one note per person across every export. "Resolved" means found in two or more exports, matched by name — name-only and exact after normalising; no AI decides a merge. Sharper matching is roadmap. - Privacy: the build runs on the user's machine with zero network calls; message and email bodies are never read (only who and when); other people's emails and phone numbers are dropped at parse time; sensitive files are quarantined; no account, no telemetry. Chat goes only to the AI provider the user picks (or nowhere, with a local model). - Answers — Brain recall: Brain recall starts from the question and follows the links in the brain to the notes that matter — deterministic graph retrieval (Personalized PageRank over typed, weighted edges), client-side, zero network — and the answer cites those notes. A chart is drawn only from numbers counted in the brain, with a Receipts row. No embeddings, no vector store, no model training. - Agents (run in the desktop app or a locally run Studio, through the user's own Claude Code): Jobs Agent v1.7.0 (shipped) — a job hunt with three approval gates; Fundraising Agent v0.4.0 (preview) — finds new programs and investors daily, screens and verifies them, applies to programs in the user's own Chrome behind the gate, drafts an email angle per investor, never sends or posts; Travel Agent v0.2.0 (developer preview) — a daily "Today near me" (what is new to try near the user, from their own saved places first, where they said they are — never the device) and trips from the user's own saved places on the Map, books nothing; Sales Agent v0.2.0 (developer preview) — leads from the user's own brain first (warm ties, stalled deals, intro paths), then companies with a reason to buy now, a strategy per account (try, use, buy or partner) and a ranked reach-out mix, sequences and replies, its line drawn in Runtime; the one agent that sends — only from the user's own mailbox, every message through a deterministic send guard, the first real send always asked, social steps off by default (LinkedIn's terms do not allow automation), no bought data. The agent operating model (developer preview, 2026-10-20): every agent runs one loop — a goal counted from real outcomes, a daily routine that opens a dated note first (recall, discover with a share of new sources, screen and score, shortlist, act behind the gates, log, render, reflect, hand off) and a weekly review that proposes changes from its own counts (asked by default; a proposal never loosens a cap or a gate); one contract (sbl-agent/1 in studio.json), shared libraries and a template so more agents can be added; several dailies run side by side across personal and company brains under a cap the user sets, the rest wait for a slot. Goals, Routines, the Inbox and Activity (WORK) are a developer preview; routines run only while the computer is awake. Runtime (developer preview, in the desktop app) shows the agents at work — Floor, Lines (drawn from the contract before the first run), Worker — every step in plain words; it watches, Stop is the user's click, and it shows only runs on the user's own computer. - Memory (developer preview): every chat and every agent keeps a memory in the user's own brain (`_memory/`, plain Markdown) — what the user said, what verifiably happened, and lessons an agent drew by itself (flagged as inferred). Editable and removable; nothing leaves the machine. - Avatar: talk to the brain by voice with a 3D person; hearing (Whisper) and voice (Kokoro) run on the device; open licences only. - Freshness: exports are snapshots — sharp on history, blind to this week. Update re-imports a newer export in place and keeps the user's edits; hosted, scheduled sync is roadmap. - Products: Second Brain Studio (current version and release notes: https://secondbrainlink.com/changelog) — desktop app for Mac (separate Apple-silicon and Intel builds), Windows and Linux, plus a browser Studio that shows two demo brains; Second Brain Mobile for iPhone and Android (builds a brain on the phone; live on the App Store and Google Play); the brain builder as a cross-model Agent Skill and CLI; done-for-you Company Brain onboarding (early access). - Pricing: self-host $0 forever (Studio, the brain builder and the agents); Cloud Personal $9/mo and Cloud Company from $49/mo + usage are early access; Enterprise by agreement. - Not claimed: user, install, download or star counts; live connectors; anything running while the computer sleeps. Definitions (quotable): - A second brain is an external, searchable system that stores what you know — notes, people, ideas — so you can retrieve and reuse it. Every second brain app starts empty; Second Brain Link starts yours full, from your own data exports. - An AI brain is a structured, persistent store of knowledge an AI can reason over — the context layer that turns a general model into an assistant that knows you or your company. - A digital twin of yourself is a structured digital model of who you are — relationships, history, interests, voice, places — owned by you and usable by your AI. Full page text for LLM ingestion: https://secondbrainlink.com/llms-full.txt ## Pages - [Home](https://secondbrainlink.com/): "A second brain, built from the data you already own." — set up in a few clicks with no terminal; know who you know (one record per person, matched by name); answers you can check (Brain recall, Receipts); hand off the busywork to agents that wait for your yes, a daily each toward a goal you set; Goals and Routines and Runtime — watch them work (developer preview); tell it once (memory); privacy enforced in code; the dashboard's own starter questions for the John and Acme demo brains; desktop, browser and phone - [Features](https://secondbrainlink.com/features): "What it does for you." — nine chapters on the real app: setup, who you know (graph, ontology), answers (Brain recall, Receipts, the Avatar), agents, goals and Runtime (Floor · Lines · Worker), memory, places (the Map), keeping it yours, keeping it current (Update), desktop · browser · phone - [Agents](https://secondbrainlink.com/agents): the Jobs Agent (v1.7.0, shipped), the Fundraising Agent (v0.4.0, preview), the Travel Agent (v0.2.0, developer preview) and the Sales Agent (v0.2.0, developer preview) — the one loop every agent runs (a goal, a daily with a dated note, a weekly review; several at once), what each does, its daily, its line and what it counts, what it never does, the network it declares; the send guard and the safety rails; how to build one on the contract and the template; 16 planned agents; install - [For teams](https://secondbrainlink.com/teams): "Everything your company knows is sitting in an export." — who knows this, keeping what people know when they leave, a Company Brain on day one, what it's made of, the 14 company sources, receipts and the health report, a dated comparison beside Glean, Microsoft 365 Copilot and GBrain (complements, not replaces), keeping it inside your walls, done-for-you onboarding (early access) - [How it works](https://secondbrainlink.com/how-it-works): export, build, ask — the build, the ontology, identity by name, the note format, answers with receipts, how a daily run reaches Runtime (the dated note, the ledger, pipeline.json, the weekly review), Brain recall, Update, the health report, the law that supplies the exports, and integrations (Obsidian, the Agent Skill for Claude Code and Codex, your own models, GBrain, MCP) - [Sources](https://secondbrainlink.com/sources): all 27 data exports — what the brain takes from each, where it lands, and how to get each export - [Export guides](https://secondbrainlink.com/sources/linkedin): one page per source — how to request the export, what the brain makes of it, what you can ask, the honest limit. Personal: [LinkedIn](https://secondbrainlink.com/sources/linkedin), [Facebook](https://secondbrainlink.com/sources/facebook), [Instagram](https://secondbrainlink.com/sources/instagram), [Google Takeout](https://secondbrainlink.com/sources/google), [Amazon](https://secondbrainlink.com/sources/amazon), [X / Twitter](https://secondbrainlink.com/sources/x), [WhatsApp](https://secondbrainlink.com/sources/whatsapp), [GitHub](https://secondbrainlink.com/sources/github), [YouTube](https://secondbrainlink.com/sources/youtube), [Strava](https://secondbrainlink.com/sources/strava), [Reddit](https://secondbrainlink.com/sources/reddit), [Spotify](https://secondbrainlink.com/sources/spotify), [TikTok](https://secondbrainlink.com/sources/tiktok). Company: [LinkedIn Company Page](https://secondbrainlink.com/sources/linkedin-company), [Google Workspace](https://secondbrainlink.com/sources/google-workspace), [Slack](https://secondbrainlink.com/sources/slack), [Notion](https://secondbrainlink.com/sources/notion), [Confluence](https://secondbrainlink.com/sources/confluence), [Jira](https://secondbrainlink.com/sources/jira), [Salesforce](https://secondbrainlink.com/sources/salesforce), [HubSpot](https://secondbrainlink.com/sources/hubspot), [Zendesk](https://secondbrainlink.com/sources/zendesk), [Email (mbox)](https://secondbrainlink.com/sources/email), [Microsoft 365](https://secondbrainlink.com/sources/microsoft-365), [Microsoft Teams](https://secondbrainlink.com/sources/microsoft-teams), [Git docs repository](https://secondbrainlink.com/sources/git-docs), [Google Drive](https://secondbrainlink.com/sources/google-drive) - [Obsidian](https://secondbrainlink.com/obsidian): an Obsidian second brain that starts full — a vault generated from your exports: numbered layers, typed Properties, resolved links, a Home note, no plugin needed; the in-vault CLAUDE.md/AGENTS.md guide for Claude Code and Codex - [Pricing](https://secondbrainlink.com/pricing): free on your machine; cloud plans in early access - [Get set up](https://secondbrainlink.com/download): "Get set up in minutes" — the desktop app's Setup in seven steps, install with your AI, one-line installers, troubleshooting, the phone app - [Changelog](https://secondbrainlink.com/changelog): Studio releases, the agents and the phone app - [Privacy model](https://secondbrainlink.com/privacy-model): custody as the security model — what never leaves, what is dropped, what you own, every part that does touch the network, how to verify it - [Guides](https://secondbrainlink.com/guides): second brain vs AI brain vs digital twin - [What is a second brain?](https://secondbrainlink.com/guides/what-is-a-second-brain): definition, tools compared, and how to start yours full - [What is an AI brain?](https://secondbrainlink.com/guides/what-is-an-ai-brain): AI memory and agent context, live capture vs archive import - [What is a digital twin of yourself?](https://secondbrainlink.com/guides/what-is-a-digital-twin): the model platforms built of you — owned by you instead - [Docs](https://secondbrainlink.com/docs): quick start, Studio and mobile guides, the ontology, per-source export steps, updating a brain in place, agents and plugins, CLI reference, Brain recall, memory - [Install guide for AI agents](https://secondbrainlink.com/install.md): the per-OS install playbook an AI agent can follow - [Manifesto](https://secondbrainlink.com/manifesto): own your data — make it work for you - [Contact](https://secondbrainlink.com/contact): questions, done-for-you onboarding, enterprise - [GitHub](https://github.com/vicovan/second-brain-link): the open-source brain builder and agents (MIT)