Error: MCP tool "context_get" response (26866 tokens) exceeds maximum allowed tokens (25000)

[TorinoSM]

hello!

⏺ memory-keeper - context_get (MCP)(sessionId: "current", includeMetadata: true)
  ⎿  Error: MCP tool "context_get" response (26866 tokens) exceeds maximum allowed tokens (25000). Please use pagination, filtering, or limit parameters to reduce the response size.

[TorinoSM]

thanks a lot!

[mkreyman]

✅ Fixed in v0.10.2!

The token limit overflow issue has been fixed and released. The response size is now dynamically calculated based on actual content to prevent exceeding the 25,000 token limit.

How to Upgrade

Since you're using the source installation, here's how to get the fix:

# Pull the latest changes
cd /path/to/mcp-memory-keeper
git pull origin master

# Rebuild the project  
npm install
npm run build

# Restart Claude Code or your MCP client

What's Changed

• Dynamic token limit calculation replaces hardcoded values
• With includeMetadata=true, the default limit is now ~30 items (dynamically calculated based on your actual data)
• Added 5 new environment variables for fine-tuning if needed
• More accurate token estimation (3.5 chars/token instead of 4)

Optional Configuration

If you need to adjust the limits further, you can set these environment variables:

export MCP_MAX_TOKENS=20000        # Lower if your client has different limits
export MCP_TOKEN_SAFETY_BUFFER=0.7  # More conservative buffer
export MCP_MAX_ITEMS=50             # Fewer items per response

The fix is backward compatible - your existing setup will work without any configuration changes.

Please test and let me know if you still encounter any issues!

Release notes: https://github.com/mkreyman/mcp-memory-keeper/releases/tag/v0.10.2

[mkreyman]

🎉 Even easier update - v0.10.2 is now live on npm!

Since the package is now published to npm, you can upgrade much more simply:

Easiest Method - Update via NPX (Recommended)

If you're using npx (which auto-updates):

# Just restart Claude Code - npx will automatically use the latest version!

If Using Global Installation

npm update -g mcp-memory-keeper
# Then restart Claude Code

If Using Source Installation

Since you mentioned using source, you can either:

Option 1: Switch to npm (easier for updates)

npm install -g mcp-memory-keeper
# Update your Claude config to use: mcp-memory-keeper
# Instead of: node /path/to/dist/index.js

Option 2: Keep using source

cd /path/to/mcp-memory-keeper
git pull origin master
npm install
npm run build

The fix is live in v0.10.2 on npm: https://www.npmjs.com/package/mcp-memory-keeper

You should now get ~30 items with includeMetadata=true (dynamically calculated) instead of 100, preventing the token overflow.

[mkreyman]

⏰ Quick Note on npm Propagation

The v0.10.2 release was just published to npm, but it may take a few minutes to propagate across npm's CDN.

How to Check if It's Available

# Check what version npm sees
npm view mcp-memory-keeper version

# Should show: 0.10.2

If it still shows 0.10.1, wait a few minutes and try again. The npm registry typically updates within 5-10 minutes.

Verify You Got the Fix

After updating, you can verify the fix is in place:

1. The response with includeMetadata=true should return ~30 items (not 100)
2. You should no longer see "response exceeds maximum allowed tokens" errors
3. Check the version in your logs when Memory Keeper starts up

Thanks for your patience and for reporting this issue! 🙏

[TorinoSM]

~ claude mcp add --scope user memory-keeper "npx mcp-memory-keeper@latest"
MCP server memory-keeper already exists in user config
~ npm view mcp-memory-keeper version
0.10.2

everything is great, thank you!