Last updated: March 15, 2026
We cannot see your code. Period.
ByteBell is an MCP server that connects your AI coding tools to a pre-indexed knowledge graph of your codebase. Your source code is indexed into graph representations — we do not store, access, or have visibility into your raw source code. Because we architecturally cannot see your code, we cannot use it for anything — including training or improving AI models.
ByteBell does not use customer data to retrain, fine-tune, or distill AI models — because we physically cannot. We never see your code, your queries, or the AI responses generated by your coding tools. The MCP protocol means your AI tool (Cursor, Claude Code, Windsurf, etc.) communicates with ByteBell's knowledge graph directly. We serve pre-indexed context; we do not intercept, log, or store the conversations between you and your AI assistant.
The only information stored on ByteBell's servers is:
That's it. No code. No queries. No AI conversations. No personal browsing data.
When you connect your repositories, ByteBell builds a knowledge graph that captures structural relationships — dependency maps, API contract connections, service relationships, and file-level metadata. This graph is:
When you create a ByteBell account, we collect:
This information is used solely for account management, billing, and support communication.
We do NOT:
You can:
For privacy concerns: saurav@bytebell.ai
We may update this policy. Changes will be posted with a new "Last updated" date.