01
Why does this matter to us?
Code, architectural decisions, tickets and documentation change continuously. Conventional context windows see only fragments and lose both history and the exact relationship to a commit.
Workshop · In internal pilot
A remotely operated code-intelligence server that indexes source code, history, documentation and business context and makes them available to AI tools through MCP.
Can an AI tool understand a large codebase precisely without relying on outdated summaries or opaque model memory?
01
Code, architectural decisions, tickets and documentation change continuously. Conventional context windows see only fragments and lose both history and the exact relationship to a commit.
02
A SHA-anchored live index combines exact symbols, a code graph and semantic search. Git webhooks update only affected parts; MCP supplies source code and evidence to Claude Code, Codex, Cursor and other clients.
03
Indexing, GitHub integration, history, MCP, OAuth, tenant isolation, connectors, backups and the operations interface are implemented and covered by automated checks. The real operator pilot is still pending.
04
Connect a real twopeaks repository in production, evaluate the most important questions against real codebases and then decide on initial customer access.