Uses too much tokens

[ilkerdemirci53]

I'm working on medium scale project. And claude code consumes all my tokens in <10 messages. It wasn't like that before. How can i optimize it?

[mrlfarano]

Token Usage Mitigation

Until this is fixed, reduce token consumption by lowering these settings in ~/.claude-mem/settings.json:

{
  "CLAUDE_MEM_OPENROUTER_MAX_CONTEXT_MESSAGES": "10",
  "CLAUDE_MEM_OPENROUTER_MAX_TOKENS": "50000",
  "CLAUDE_MEM_CONTEXT_OBSERVATIONS": "25",
  "CLAUDE_MEM_CONTEXT_FULL_COUNT": "3",
  "CLAUDE_MEM_CONTEXT_SESSION_COUNT": "5"
}

Why this helps:

• Each failed retry consumes ~10k tokens
• MAX_CONTEXT_MESSAGES: "10" (down from 20) cuts token usage per attempt in half
• Smaller context settings reduce the payload size

For users on free tiers: Use a free model to avoid costs during the crash loop:

{
  "CLAUDE_MEM_OPENROUTER_MODEL": "openbmb/mimo-v2-flash:free"
}

Monitor your usage: Check the logs for token consumption:

grep "API usage" ~/.claude-mem/logs/claude-mem-$(date +%Y-%m-%d).log

[ilkerdemirci53]

Token Usage Mitigation

Until this is fixed, reduce token consumption by lowering these settings in ~/.claude-mem/settings.json:

{
  "CLAUDE_MEM_OPENROUTER_MAX_CONTEXT_MESSAGES": "10",
  "CLAUDE_MEM_OPENROUTER_MAX_TOKENS": "50000",
  "CLAUDE_MEM_CONTEXT_OBSERVATIONS": "25",
  "CLAUDE_MEM_CONTEXT_FULL_COUNT": "3",
  "CLAUDE_MEM_CONTEXT_SESSION_COUNT": "5"
}

Why this helps:

• Each failed retry consumes ~10k tokens
• MAX_CONTEXT_MESSAGES: "10" (down from 20) cuts token usage per attempt in half
• Smaller context settings reduce the payload size

For users on free tiers: Use a free model to avoid costs during the crash loop:

{
  "CLAUDE_MEM_OPENROUTER_MODEL": "openbmb/mimo-v2-flash:free"
}

Monitor your usage: Check the logs for token consumption:

grep "API usage" ~/.claude-mem/logs/claude-mem-$(date +%Y-%m-%d).log

Thanks, but if this Claude-mem burns through my tokens much faster (not 20x more usage, but 20x less 😅), why is it still giving wrong information and wasting my time?