BaseThread

About BaseThread

BaseThread is the team-scoped version of the shared-AI-context idea: instead of giving one person's assistant a memory, it gives a whole company one context graph that every teammate's AI reads from and writes back to. The graph is organized into five explicit layers. Company holds mission, brand voice, product strategy, target personas, competitors, roadmap, customer base, and tech stack. Products holds product decisions, active bets, pricing, positioning, OKRs, architecture decisions, release notes, feature specs, and integrations. Teams holds members, assigned tasks, process and tooling decisions, goals, and pipeline decisions. Projects holds briefs, success metrics, deliverables, owners, open questions, risks, and launch checklists. You holds your role, preferences, 1:1 notes, current focus, and working hours. The mechanism that keeps it current is the interesting part: each member's local AI acts as scribe, summarizing what it just did and calling `ledger.activity`, `ledger.decide`, and `ledger.task` to write activity, decisions with rationale, and next steps into the BaseThread cloud, which harmonizes them across members and pushes the result back to everyone's tools as three streams. So a question like "what did we decide about the pricing model, and why" gets answered from the ledger with the rationale and the person who made the call attached. It runs over both local and remote MCP, so Claude Code, Cursor, ChatGPT, and any MCP client can read and write it. Multi-tier permissions let each layer's owner hold source of truth.

Key Features

Tags

More AI Agent Infrastructure tools