What problem does this solve?
I use a custom tool to concatenate all my codebase into a text file to paste into ChatGPT/DeepSeek/Qwen to draft an initial implementation plan before running agents — saves tokens and improves planning accuracy for initial draft.
Currently: get_architecture returns JSON (great for agents, not for LLM context). I manually copy files → lose import relationships → LLM makes assumptions.
Need: A command that exports the indexed codebase as plain text I can copy-paste into chat LLMs.
Proposed solution
Add an MCP tool export_codebase that outputs structured text:
codebase-memory-mcp cli export_codebase '{"project":"myapp"}'
Alternatives considered
JSON-only export — Could reuse get_architecture as-is, but JSON doesn't paste well into chat interfaces; text is more natural.
Markdown format — Possible as a format option, but text is simpler and more portable.
Server-side compression — Could gzip large exports, but text is already token-efficient; not needed.
Confirmations
What problem does this solve?
I use a custom tool to concatenate all my codebase into a text file to paste into ChatGPT/DeepSeek/Qwen to draft an initial implementation plan before running agents — saves tokens and improves planning accuracy for initial draft.
Currently: get_architecture returns JSON (great for agents, not for LLM context). I manually copy files → lose import relationships → LLM makes assumptions.
Need: A command that exports the indexed codebase as plain text I can copy-paste into chat LLMs.
Proposed solution
Add an MCP tool export_codebase that outputs structured text:
Alternatives considered
JSON-only export — Could reuse get_architecture as-is, but JSON doesn't paste well into chat interfaces; text is more natural.
Markdown format — Possible as a format option, but text is simpler and more portable.
Server-side compression — Could gzip large exports, but text is already token-efficient; not needed.
Confirmations