Every team solves the same problems over and over. A production outage gets debugged, the fix lands, and six months later someone else hits the exact same issue — starting from scratch. The person who fixed it last time may have left the project entirely.
workthin breaks that cycle. When someone solves a problem, workthin records the solution as structured knowledge — problem, environment, what was tried, and what worked. Next time anyone encounters the same class of problem, their AI tool finds the answer automatically.
How it works
workthin runs as an MCP server that connects to Claude Code, Claude Desktop, ChatGPT, Cursor, VS Code / GitHub Copilot, Codex CLI, GitHub Copilot CLI, Gemini CLI, and any other MCP-compatible tool. Setup is one command:
npx workthin@latest initnpx workthin@latest init handles authentication, detects your AI tool, configures the MCP connection, and generates /workthin (record) and /workthin-search (search) commands. From that point, your AI tool can search and create knowledge directly within conversations.
Three scopes
Knowledge lives in one of three scopes:
- personal — only you can see it
- project — shared with your project team (auto-detected from git remote)
- global — shared with all workthin users
When you create knowledge without specifying a scope, workthin saves it as personal and suggests a more appropriate scope. Over time, individual solutions become team knowledge, then community knowledge.
Security by default
A 3-stage Protection Layer masks secrets before anything is stored: regex pattern matching, Shannon entropy analysis, and AI classification. API keys, tokens, and credentials are caught and replaced with [MASKED:pattern_name] automatically.
Pricing
Free plan includes everything — unlimited storage, CLI, MCP, semantic search, and secret protection. The only limit is 30 global reads per month. Pro ($10/month) removes that limit.
workthin is available now at workthin.com.