The installed CLI is the command source of truth. Use slm --help and slm <command> --help; commands that advertise --json provide structured output.

slm remember "Fixed the auth bug — JWT expiry was set to 1 hour instead of 24"
slm remember "API rate limit is 100/min" --tags "api,config"
slm remember "Important fact" --json    # Agent-native JSON output
slm remember "Shared decision" --scope shared --shared-with team-a
slm remember "Wait for enrichment" --sync --json

Store a memory. The default daemon path commits raw evidence plus a queryable SQLite relational/FTS projection, then enriches the same durable operation in the background.

Options:

• --tags "tag1,tag2" — Add tags
• --json — Output structured JSON (for agents, scripts, CI/CD)
• --sync — Wait for all declared derivation and projector stages
• --scope personal|shared|global — Set visibility
• --shared-with "profile-a,profile-b" — Name readers for shared scope

JSON output includes operation_id, materialization_state, and fact IDs. If the daemon cannot start, raw evidence enters the legacy offline spool; replay submits it through M018 before marking that spool row done.

slm recall "JWT token configuration"
slm recall "auth setup" --limit 5 --json

Retrieve memories using the candidate producers healthy in the configured mode, followed by fusion, optional reranking, and graph-based score enhancement. Results follow Score Contract v2.

Options:

• --limit N — Number of results (default: 10)
• --json — Output structured JSON

slm list                    # Last 20 memories (shows IDs for delete/update)
slm list -n 50 --json       # JSON output with fact IDs

List recent memories chronologically. Shows fact IDs needed for delete and update operations.

Options:

• --limit N / -n N — Number of entries (default: 20)
• --json — Output structured JSON

slm trace "JWT token configuration"
slm trace "database port" --json

Same as recall, but shows per-channel score breakdown. Current candidate producers are dense semantic, BM25 lexical, temporal, Hopfield associative, and spreading activation. Entity-graph information can enhance a post-fusion score but is not a separate candidate producer.

Options:

• --json — Output structured JSON with channel_scores per result

slm forget "JWT token configuration"
slm forget "old staging config" --yes         # Skip confirmation
slm forget "old stuff" --json                 # Preview matches (no delete)
slm forget "old stuff" --json --yes           # Delete and return JSON

Delete memories matching a query. Shows matching memories and asks for confirmation before deleting.

Options:

• --yes / -y — Skip confirmation prompt
• --json — Output structured JSON (without --yes: preview only; with --yes: delete and confirm)

slm delete <fact_id>                # Delete by exact ID (with confirmation)
slm delete <fact_id> --yes          # Skip confirmation
slm delete <fact_id> --json --yes   # Delete and return JSON

Delete a specific memory by exact fact ID. Use slm list to find fact IDs.

Options:

• --yes / -y — Skip confirmation prompt
• --json — Output structured JSON

slm update <fact_id> "corrected content"
slm update <fact_id> "new text" --json

Update the content of a specific memory. Use slm list to find fact IDs.

Options:

• --json — Output structured JSON

slm connect        # Auto-detect and configure all installed IDEs
slm connect --list # Show which IDEs are configured
slm mcp            # Start MCP server (stdio transport — used by IDEs)

The slm mcp command is what your IDE calls internally. You typically don't run it directly — your IDE's MCP config handles it:

{
  "mcpServers": {
    "superlocalmemory": {
      "command": "slm",
      "args": ["mcp"]
    }
  }
}

slm profile list              # List all profiles
slm profile create <name>     # Create a new profile
slm profile switch <name>     # Switch active profile

Personal facts are profile-isolated by default. Shared and global recall are opt-in and remain subject to the configured scope policy; do not use profiles as a substitute for operating-system or tenant isolation.

slm migrate                   # Upgrade V2 database to V3
slm migrate --rollback        # Undo migration

slm dashboard                 # Open web dashboard at http://localhost:8765
slm dashboard --port 9000     # Use a custom port

17-tab dashboard: memory browser, knowledge graph, recall lab, trust scores, math health, compliance, learning, IDE connections, settings, and more.

# Store a decision with tags
slm remember "Chose PostgreSQL over MongoDB for the user service. Reason: ACID transactions needed for billing." --tags "architecture,database"

# Recall with channel breakdown
slm trace "database decision for user service"

# Check system status
slm status

# Check math layer health
slm health

# Switch to full power mode
slm mode c

# Open the dashboard
slm dashboard

Commands that advertise --json provide structured output. Recall fields keep ranking relevance separate from stored-memory confidence:

{
  "success": true,
  "command": "recall",
  "version": "<installed-version>",
  "data": {
    "results": [
      {
        "fact_id": "abc123",
        "content": "Database uses PostgreSQL 16",
        "relevance_score": 0.87,
        "ranking_score": 0.0132,
        "memory_confidence": 0.7,
        "rank_position": 1
      }
    ],
    "count": 1,
    "score_contract_version": "2",
    "calibration_status": "uncalibrated",
    "answer_confidence": null
  },
  "next_actions": [
    {"command": "slm list --json", "description": "List recent memories"}
  ]
}

Structured-output support is explicit per command and can expand between releases.

Usage with jq:

slm recall "auth" --json | jq '.data.results[0].content'
slm list --json | jq '.data.results[].fact_id'
slm status --json | jq '.data.mode'

In CI/CD (GitHub Actions):

- name: Store deployment info
  run: slm remember "Deployed ${{ github.sha }}" --json

- name: Verify memory health
  run: slm status --json | jq -e '.success'

SuperLocalMemory exposes both MCP and CLI surfaces:

This is an orientation list, not the complete installed surface. Run slm --help for the installed release.

Part of Qualixar | Created by Varun Pratap Bhardwaj