Configuration

Settings are stored in ~/.claude-mem/settings.json. The file is created automatically with defaults on first run. You can edit it directly or use the Context Settings Modal at http://localhost:37777 for live preview of changes.

Core settings

Setting	Default	Description
`CLAUDE_MEM_MODEL`	`sonnet`	AI model for processing observations when using the Claude provider
`CLAUDE_MEM_PROVIDER`	`claude`	AI provider: `claude`, `gemini`, or `openrouter`
`CLAUDE_MEM_MODE`	`code`	Active mode profile (e.g. `code--es`, `email-investigation`)
`CLAUDE_MEM_WORKER_PORT`	`37777`	Port for the background worker service
`CLAUDE_MEM_WORKER_HOST`	`127.0.0.1`	Host address for the worker service
`CLAUDE_MEM_SKIP_TOOLS`	`ListMcpResourcesTool,SlashCommand,Skill,TodoWrite,AskUserQuestion`	Comma-separated list of tools excluded from observation capture

Available Claude models

Shorthand model names automatically resolve to the latest version:

Shorthand	Description
`haiku`	Fast and cost-efficient
`sonnet`	Balanced performance (default)
`opus`	Most capable

Skipping tools

You can control which tool executions are captured as observations. Edit ~/.claude-mem/settings.json:

{
  "CLAUDE_MEM_SKIP_TOOLS": "ListMcpResourcesTool,SlashCommand,Skill"
}

Common customizations:

Include TodoWrite: Remove it from the skip list to track task planning in your memory
Include AskUserQuestion: Remove it to capture decision-making conversations
Skip additional tools: Add tool names to reduce noise in your observations

Changes take effect on the next tool execution — no worker restart needed.

Context injection settings

Claude Mem injects past observations into each new session. The Context Settings Modal at http://localhost:37777 is the recommended way to configure these settings — it provides a live terminal preview showing exactly what will be injected at the start of your next session.

Open the web viewer

Navigate to http://localhost:37777 in your browser.

Open settings

Click the gear icon in the header. The settings panel opens on the right.

Adjust and preview

Change any setting and see the Terminal Preview on the left update in real time.

Auto-save

Settings are saved automatically as you make changes — no submit button needed.

Loading settings

Control how many observations are injected per session:

Setting	Default	Range	Description
`CLAUDE_MEM_CONTEXT_OBSERVATIONS`	`50`	1–200	Total number of recent observations to include
`CLAUDE_MEM_CONTEXT_SESSION_COUNT`	`10`	1–50	Number of recent sessions to pull observations from

Higher values give Claude more historical context but increase session startup time and token usage. The default of 50 observations from 10 sessions balances context richness with performance.

Filter settings

Control which observation types and concepts are included. Types — select any combination:

Value	Description
`bugfix`	Bug fixes and error resolutions
`feature`	New functionality additions
`refactor`	Code restructuring
`discovery`	Learnings about how code works
`decision`	Architectural or design decisions
`change`	General code changes

Concepts — select any combination:

Value	Description
`how-it-works`	System behavior explanations
`why-it-exists`	Rationale for code or design
`what-changed`	Change summaries
`problem-solution`	Problem and solution pairs
`gotcha`	Edge cases and pitfalls
`pattern`	Recurring patterns
`trade-off`	Design trade-offs

Use the All or None buttons in the modal to quickly select or deselect all options. Setting these manually in ~/.claude-mem/settings.json:

{
  "CLAUDE_MEM_CONTEXT_OBSERVATION_TYPES": "bugfix,decision,discovery",
  "CLAUDE_MEM_CONTEXT_OBSERVATION_CONCEPTS": "how-it-works,gotcha,problem-solution"
}

Display settings

Control how observations appear in the injected context:

Setting	Default	Options	Description
`CLAUDE_MEM_CONTEXT_FULL_COUNT`	`5`	0–20	How many observations show their full expanded details
`CLAUDE_MEM_CONTEXT_FULL_FIELD`	`narrative`	`narrative`, `facts`	Which field is expanded in full observations

The most recent N observations (set by Count) show their full narrative or facts. The remaining observations appear in a compact table format showing only title, type, and token counts. Token economics toggles:

Setting	Default	Description
`CLAUDE_MEM_CONTEXT_SHOW_READ_TOKENS`	`true`	Show tokens required to read each observation
`CLAUDE_MEM_CONTEXT_SHOW_WORK_TOKENS`	`true`	Show tokens invested in creating the observation
`CLAUDE_MEM_CONTEXT_SHOW_SAVINGS_AMOUNT`	`true`	Show total tokens saved by reusing cached context
`CLAUDE_MEM_CONTEXT_SHOW_SAVINGS_PERCENT`	`false`	Show savings as a percentage instead of absolute token count
`CLAUDE_MEM_CONTEXT_SHOW_TERMINAL_OUTPUT`	`false`	Include terminal/command output snippets in context display

Advanced context settings

Setting	Default	Description
`CLAUDE_MEM_CONTEXT_SHOW_LAST_SUMMARY`	`false`	Include the previous session’s summary in injected context
`CLAUDE_MEM_CONTEXT_SHOW_LAST_MESSAGE`	`false`	Include the previous session’s final message in injected context

Full manual configuration example

{
  "CLAUDE_MEM_CONTEXT_OBSERVATIONS": "100",
  "CLAUDE_MEM_CONTEXT_SESSION_COUNT": "20",
  "CLAUDE_MEM_CONTEXT_OBSERVATION_TYPES": "bugfix,decision,discovery",
  "CLAUDE_MEM_CONTEXT_OBSERVATION_CONCEPTS": "how-it-works,gotcha",
  "CLAUDE_MEM_CONTEXT_FULL_COUNT": "10",
  "CLAUDE_MEM_CONTEXT_FULL_FIELD": "narrative",
  "CLAUDE_MEM_CONTEXT_SHOW_READ_TOKENS": "true",
  "CLAUDE_MEM_CONTEXT_SHOW_WORK_TOKENS": "true",
  "CLAUDE_MEM_CONTEXT_SHOW_SAVINGS_AMOUNT": "true",
  "CLAUDE_MEM_CONTEXT_SHOW_LAST_SUMMARY": "false",
  "CLAUDE_MEM_CONTEXT_SHOW_LAST_MESSAGE": "false"
}

Gemini provider settings

Use Google Gemini as the AI provider for observation processing. A free tier is available via Google AI Studio.

Setting	Default	Description
`CLAUDE_MEM_GEMINI_API_KEY`	—	Your Gemini API key (get a free key at aistudio.google.com)
`CLAUDE_MEM_GEMINI_MODEL`	`gemini-2.5-flash-lite`	Model to use: `gemini-2.5-flash-lite`, `gemini-2.5-flash`, `gemini-3-flash-preview`

To activate Gemini:

{
  "CLAUDE_MEM_PROVIDER": "gemini",
  "CLAUDE_MEM_GEMINI_API_KEY": "your-api-key-here",
  "CLAUDE_MEM_GEMINI_MODEL": "gemini-2.5-flash-lite"
}

OpenRouter provider settings

Use OpenRouter to access 100+ models for observation processing.

Setting	Default	Description
`CLAUDE_MEM_OPENROUTER_API_KEY`	—	Your OpenRouter API key (get one at openrouter.ai/keys)
`CLAUDE_MEM_OPENROUTER_MODEL`	`xiaomi/mimo-v2-flash:free`	Model identifier — supports 100+ models
`CLAUDE_MEM_OPENROUTER_MAX_CONTEXT_MESSAGES`	`20`	Maximum messages in conversation history
`CLAUDE_MEM_OPENROUTER_MAX_TOKENS`	`100000`	Token budget safety limit
`CLAUDE_MEM_OPENROUTER_SITE_URL`	—	Optional URL for OpenRouter analytics
`CLAUDE_MEM_OPENROUTER_APP_NAME`	`claude-mem`	Optional app name for OpenRouter analytics

To activate OpenRouter:

{
  "CLAUDE_MEM_PROVIDER": "openrouter",
  "CLAUDE_MEM_OPENROUTER_API_KEY": "your-api-key-here",
  "CLAUDE_MEM_OPENROUTER_MODEL": "xiaomi/mimo-v2-flash:free"
}

System configuration

Setting	Default	Description
`CLAUDE_MEM_DATA_DIR`	`~/.claude-mem`	Data directory location
`CLAUDE_MEM_LOG_LEVEL`	`INFO`	Log verbosity: `DEBUG`, `INFO`, `WARN`, `ERROR`, or `SILENT`
`CLAUDE_MEM_PYTHON_VERSION`	`3.13`	Python version for chroma-mcp
`CLAUDE_MEM_EXCLUDED_PROJECTS`	—	Comma-separated project names to exclude from memory capture
`CLAUDE_MEM_MAX_CONCURRENT_AGENTS`	—	Maximum number of concurrent AI processing agents in the worker
`CLAUDE_CODE_PATH`	(auto-detect)	Path to the Claude Code CLI (primarily needed on Windows)

Data directory structure

~/.claude-mem/
├── claude-mem.db           # SQLite database
├── settings.json           # User settings
├── .install-version        # Cached version for smart installer
├── worker.port             # Current worker port file
└── logs/
    ├── worker-out.log      # Worker stdout logs
    └── worker-error.log    # Worker stderr logs

In production (installed plugin), the data directory is always ~/.claude-mem/ regardless of CLAUDE_PLUGIN_ROOT. Use CLAUDE_MEM_DATA_DIR to override during development only.

Worker service management

The worker service is a background process managed by Bun. It starts automatically the first time you open a Claude Code session and continues running in the background.

npm run worker:start

Custom worker port

If port 37777 is already in use, configure a different port:

{
  "CLAUDE_MEM_WORKER_PORT": "38000"
}

Then restart the worker:

npm run worker:restart

Verify the new port is active:

cat ~/.claude-mem/worker.port

Debug logging

Enable verbose logging for the worker:

export DEBUG=claude-mem:*
npm run worker:restart
npm run worker:logs

Folder context files

Claude Mem can automatically generate CLAUDE.md files in your project folders containing activity timelines. This feature is disabled by default.

Setting	Default	Description
`CLAUDE_MEM_FOLDER_CLAUDEMD_ENABLED`	`false`	Enable auto-generation of folder `CLAUDE.md` files

Enable it in ~/.claude-mem/settings.json:

{
  "CLAUDE_MEM_FOLDER_CLAUDEMD_ENABLED": "true"
}

Hooks configuration

Hooks are defined in plugin/hooks/hooks.json. They fire at each stage of the Claude Code session lifecycle.

{
  "description": "Claude-mem memory system hooks",
  "hooks": {
    "Setup": [{
      "matcher": "*",
      "hooks": [{
        "type": "command",
        "command": "_R=\"${CLAUDE_PLUGIN_ROOT}\"; [ -z \"$_R\" ] && _R=\"$HOME/.claude/plugins/marketplaces/thedotmack/plugin\"; \"$_R/scripts/setup.sh\"",
        "timeout": 300
      }]
    }],
    "SessionStart": [{
      "matcher": "startup|clear|compact",
      "hooks": [
        {
          "type": "command",
          "command": "_R=\"${CLAUDE_PLUGIN_ROOT}\"; [ -z \"$_R\" ] && _R=\"$HOME/.claude/plugins/marketplaces/thedotmack/plugin\"; node \"$_R/scripts/smart-install.js\"",
          "timeout": 300
        },
        {
          "type": "command",
          "command": "_R=\"${CLAUDE_PLUGIN_ROOT}\"; [ -z \"$_R\" ] && _R=\"$HOME/.claude/plugins/marketplaces/thedotmack/plugin\"; node \"$_R/scripts/bun-runner.js\" \"$_R/scripts/worker-service.cjs\" start",
          "timeout": 60
        },
        {
          "type": "command",
          "command": "_R=\"${CLAUDE_PLUGIN_ROOT}\"; [ -z \"$_R\" ] && _R=\"$HOME/.claude/plugins/marketplaces/thedotmack/plugin\"; node \"$_R/scripts/bun-runner.js\" \"$_R/scripts/worker-service.cjs\" hook claude-code context",
          "timeout": 60
        }
      ]
    }],
    "UserPromptSubmit": [{
      "hooks": [{
        "type": "command",
        "command": "_R=\"${CLAUDE_PLUGIN_ROOT}\"; [ -z \"$_R\" ] && _R=\"$HOME/.claude/plugins/marketplaces/thedotmack/plugin\"; node \"$_R/scripts/bun-runner.js\" \"$_R/scripts/worker-service.cjs\" hook claude-code session-init",
        "timeout": 60
      }]
    }],
    "PostToolUse": [{
      "matcher": "*",
      "hooks": [{
        "type": "command",
        "command": "_R=\"${CLAUDE_PLUGIN_ROOT}\"; [ -z \"$_R\" ] && _R=\"$HOME/.claude/plugins/marketplaces/thedotmack/plugin\"; node \"$_R/scripts/bun-runner.js\" \"$_R/scripts/worker-service.cjs\" hook claude-code observation",
        "timeout": 120
      }]
    }],
    "Stop": [{
      "hooks": [{
        "type": "command",
        "command": "_R=\"${CLAUDE_PLUGIN_ROOT}\"; [ -z \"$_R\" ] && _R=\"$HOME/.claude/plugins/marketplaces/thedotmack/plugin\"; node \"$_R/scripts/bun-runner.js\" \"$_R/scripts/worker-service.cjs\" hook claude-code summarize",
        "timeout": 120
      }]
    }],
    "SessionEnd": [{
      "hooks": [{
        "type": "command",
        "command": "node -e \"...\" ",
        "timeout": 5
      }]
    }]
  }
}

Hook timeout recommendations

Hook	Timeout	Reason
`Setup`	300s	First-time dependency installation
`SessionStart` (smart-install)	300s	Dependency check (fast ~10ms when cached)
`SessionStart` (worker start)	60s	Start background worker service
`SessionStart` (context inject)	60s	Retrieve and inject session context
`UserPromptSubmit`	60s	Lightweight session initialization
`PostToolUse`	120s	Observation capture and queuing
`Stop`	120s	Summary generation
`SessionEnd`	5s	Mark session complete (fire-and-forget)

The SessionStart smart-install step is fast (~10ms) when dependencies are already installed. It only runs the full install on version changes.

Mode configuration

Modes configure the active workflow profile and language for observation processing.

{
  "CLAUDE_MEM_MODE": "code"
}

Examples:

{
  "CLAUDE_MEM_MODE": "code--es"
}

Version channel

Switch between stable and beta versions from the web viewer at http://localhost:37777.

Open settings

Click the gear icon in the viewer header.

Find version channel

Locate the Version Channel section in the settings panel.

Switch versions

Click Try Beta (Endless Mode) for experimental features, or Switch to Stable to return to the production release.

Your memory data is preserved when switching versions — only the plugin code changes.

Endless Mode is experimental and slower than standard mode.

Configuration best practices

Use defaults where possible

The default configuration works well for most use cases. Start with defaults and only change settings when you have a specific reason.

Override selectively

Set only what you need in ~/.claude-mem/settings.json. Unset values automatically fall back to defaults.

Restart worker after changes

Most settings require a worker restart to take effect:

npm run worker:restart

Exception: CLAUDE_MEM_SKIP_TOOLS changes take effect immediately on the next tool execution.

Monitor logs after changes

After restarting the worker, check logs to confirm the configuration was applied correctly:

npm run worker:logs

Get Started

Using Claude Mem

AI Providers

Best Practices

Configuration & Development

Core settings

Available Claude models

Skipping tools

Context injection settings

Loading settings

Filter settings

Display settings

Advanced context settings

Full manual configuration example

Gemini provider settings

OpenRouter provider settings

System configuration

Data directory structure

Worker service management

Custom worker port

Debug logging

Folder context files

Hooks configuration

Hook timeout recommendations

Mode configuration

Version channel

Configuration best practices

Build docs developers (and LLMs) love

Get Started

Using Claude Mem

AI Providers

Best Practices

Configuration & Development

​Core settings

​Available Claude models

​Skipping tools

​Context injection settings

​Accessing the Context Settings Modal

​Loading settings

​Filter settings

​Display settings

​Advanced context settings

​Full manual configuration example

​Gemini provider settings

​OpenRouter provider settings

​System configuration

​Data directory structure

​Worker service management

​Custom worker port

​Debug logging

​Folder context files

​Hooks configuration

​Hook timeout recommendations

​Mode configuration

​Version channel

​Configuration best practices

Build docs developers (and LLMs) love

Core settings

Available Claude models

Skipping tools

Context injection settings

Accessing the Context Settings Modal

Loading settings

Filter settings

Display settings

Advanced context settings

Full manual configuration example

Gemini provider settings

OpenRouter provider settings

System configuration

Data directory structure

Worker service management

Custom worker port

Debug logging

Folder context files

Hooks configuration

Hook timeout recommendations

Mode configuration

Version channel

Configuration best practices