# BaseThread

> Shared context for every AI tool your team uses. Activity, decisions, tasks, and the content across all five layers are the glue that keeps every teammate's AI current and as smart as it can be. Define your company, products, teams, and projects once; every tool reads it over MCP and writes back, so the whole team's AI stays on the same page and gets sharper every day.

## Overview

BaseThread is a team second brain: a shared context graph your AI writes to as it works and every AI tool reads from, kept current automatically. Unlike a personal second brain that one person maintains by hand (Notion, Obsidian), BaseThread is team-shared and AI-maintained. A team curates its company, products, teams, projects, decisions, and people once, in one structured place, and every member's AI keeps it current as it works. Every member's AI tools (Claude Code, Cursor, Gemini CLI, Windsurf, GitHub Copilot, ChatGPT) read the right slice of that context at the start of every session over MCP, and write back activity, decisions, and tasks as work happens, so no one re-explains the same context twice and the shared layer coordinates what's next.

BaseThread's context is AI-written first. BaseThread does not connect to your tools; your AI already does. Your AI reads the tools where work happens (Notion, HubSpot, Atlassian, Slack, and your docs) with the access it already has and writes their signal into the graph, scoped to the right node and additive to the AI-written streams. It is the signal, not a raw index. The graph is custom context you collaborate on and curate across the levels, plus the harmonization of activity, decisions, and tasks back to each member's local AI. That harmonization is the core.

## What makes it different

BaseThread is the only shared-context layer that is all four at once: team-shared (not personal, single-tool memory), cross-tool (read and written by every AI tool over MCP), AI-witnessed (your AI writes the record automatically, no manual logging), and company-structured + harmonized (organized by your real hierarchy and reconciled across the whole team). And with tasks, it does not just remember the work, it coordinates it: delegate, accept, and track across tools.

## The layers

A hierarchy managed as one system. For teams there are five levels: Company, Products, Teams, Projects, and You, with initiatives grouping related projects into one program when work is bigger than a single project. For individuals there are three: Me, Family, and Projects. Context defined higher up is inherited downward, and each tool reads only the slice relevant to the task, scoped by access.

## Activity, decisions & tasks (the core)

There are three living streams. Activity is the high-frequency record of what each teammate (and their AI) shipped, changed, and ran. Decisions are the low-frequency record of what was settled, with the rationale that produced each call. Tasks are the forward-looking stream of what's next and who owns it, assignable and trackable from any tool. All three are written and read by each member's local AI as work happens: the AI summarizes activity at session end, extracts decisions when it recognizes one, and can create, assign, and update tasks in place. Entries land at the level they belong to (Company, Products, Teams, Projects, You), are harmonized in the BaseThread cloud, and are pushed back to every member's local environment. You can ask what shipped this week, what was decided about something months ago, and what's on your plate right now, and any agent answers all three. No manual logging, and no raw dumps. Your AI can also read the tools where work already happens, like Notion, HubSpot, Atlassian, and Slack, with the access it already has, and write their decisions, context, and activity into the graph, scoped to the right place, additive to the AI-written streams. BaseThread never connects to those tools; your AI is the integration.

## Two surfaces

- Web platform: where teams collaborate on, maintain, and manage the source of truth (the layers, projects, members, and decisions).
- Two ways to connect AI tools: (1) the BaseThread Mac app runs a local MCP bridge for desktop clients (Claude Code, Cursor, Windsurf, etc.); (2) remote MCP at mcp.basethread.ai for any client anywhere (ChatGPT desktop, Claude desktop, hosted agents). Same shared team context either way.

## Works with your tools

BaseThread does not connect to your tools. Your AI already does. The AI tools your team uses (Claude, Cursor, ChatGPT, and any MCP client) already hold the connections to Notion, Atlassian (Jira and Confluence), Slack, HubSpot, Linear, GitHub, and your docs. BaseThread is the shared brain they read before working and write to after. As your AI works, it writes the signal, the decisions, context, and activity, into the right place in your graph, scoped to the team or project you choose. It is not enterprise search: BaseThread never indexes your tools into a box you query. Your brain stays current because your team works through it, every session reading before and writing after.

## Key features

- A team second brain, maintained by your AI, not by hand
- A shared context graph every AI tool reads automatically over MCP
- Your AI writes and updates the context as it works (not just reads it), focused on the area you are working in
- People and 1:1 notes: per-person relationship memory for members and external contacts
- State digest: an auto-refreshed "where this stands" synthesis per team, project, or workspace
- Initiatives: group related projects into one program, a tier above projects
- Three streams (activity, decisions, and tasks) written by each member's local AI as work happens, harmonized to every member's tools
- Shared tasks: assign, accept, and track work across every AI tool, not just remember it
- Five-level hierarchy for teams (Company, Products, Teams, Projects, You); three for individuals
- Cross-role: engineering, design, product, go-to-market, and cross-functional teams share one project brief
- Local-first sync with fast offline reads
- Encrypted at rest, local-first, never used to train models
- Works with everything your AI can reach: your AI reads Notion, HubSpot, Atlassian, Slack, and your docs and writes the signal into your shared brain, scoped to the right place, never a raw index
- Works with Claude Code, Cursor, Gemini CLI, Windsurf, GitHub Copilot, ChatGPT
- Personal mode for individuals and families

## Solutions

By role: Engineering, Design, Product, and Go-to-market. By team shape: Cross-functional teams. Plus Personal use for individuals and families.

## Pricing

Freemium. Free covers teams of up to 5 with every feature and 30 days of history. Team is $5 per user per month for unlimited seats and unlimited history. Enterprise is custom, adding SSO/SAML, SCIM, audit logs, custom roles, data residency, on-prem, and a DPA. There is no Personal Pro tier. The core loop (activity, decisions, and tasks) is included on every plan; quantity limits drive upgrades.

## Compatibility

macOS (the native agent) plus any MCP-capable AI tool: Claude Code, Cursor, Gemini CLI, Windsurf, GitHub Copilot, and ChatGPT. The web platform runs in any modern browser.

## Links

- Website: https://basethread.ai
- Pricing: https://basethread.ai/pricing
- Blog: https://basethread.ai/blog