Grainulator is a Claude Code plugin for structured research sprints. It runs entirely on your local machine as a set of CLI tools and MCP servers. There is no Grainulator backend, no cloud service, and no user account.
None. Grainulator does not collect, transmit, or store any user data on remote servers. All sprint data (claims, compilations, output artifacts) is stored locally in your project root (claims.json, compilation.json, output/).
Grainulator bundles three local MCP servers that run as stdio processes on your machine:
claims.json and compilation.json locally)These servers never contact external endpoints. All data they process stays on your machine.
Grainulator includes an optional connection to DeepWiki (mcp.deepwiki.com), a third-party HTTP MCP server that provides documentation for open-source projects. When you use DeepWiki-powered tools, your queries about open-source repositories are sent to DeepWiki's servers.
DeepWiki is operated by a third party, not by Grainulator. You can remove the deepwiki entry from .mcp.json to disable this connection entirely. All other Grainulator functionality continues to work without it.
Grainulator itself does not require or store API keys. Claude Code (the host environment) uses your Anthropic API key, which is managed by Claude Code. See Anthropic's privacy policy.
All Grainulator data lives in your local filesystem:
claims.json, compilation.json, output/ in your project root~/.claude/plugins/cache/CLAUDE_PLUGIN_DATA/silo/We have no access to any of this data. You control it entirely. Delete it whenever you want.
Grainulator is a developer tool distributed via npm and Claude Code. It is not directed at children under 13.
If we change this policy, we will update this page and the effective date above. Since we collect no data, we do not anticipate material changes.
Questions? Email info@grainulator.app or open an issue at github.com/grainulation/grainulator.