// SCAN REPORT

Mapi Agent Memory

283 tools. 114 can modify or destroy data without limits.

4 destructive tools with no built-in limits. Policy required.

Last updated: 5 Oct 2026

114 can modify or destroy data

169 read-only

283 tools total

Community server · catalogue entry checked 05/10/2026

How to control Mapi Agent Memory ↓

TOOL MAP

What Mapi Agent Memory exposes to your agents

Read (169) Write / Execute (110) Destructive / Financial (4)

Running more than one server? Check your whole stack →

MOST DANGEROUS TOOLS

Critical Risk

The most dangerous Mapi Agent Memory tools

Destructive · Critical delete_path The name 'delete_path' strongly suggests deletion of files or data paths without reversibility. Destructive · High rollback_memory_pointer_lifecycle_remediation_execution A rollback operation reverts or undoes previously applied remediation changes. Destructive · High expire_memory_capture_item Expiring a durable capture review item permanently removes it from the active review pipeline without producing any compensating artifact (memory). Destructive · High gemma_worker_cancel_job gemma_worker_cancel_job

114 of Mapi Agent Memory's 283 tools can modify, destroy, or commit something on every call — and an agent calls them with no built-in limits.

POLICY

How to control Mapi Agent Memory

PolicyLayer is an MCP gateway — it sits between your AI agents and Mapi Agent Memory, and nothing reaches the server without passing your rules. These are the rules we recommend:

Deny destructive operations

{
  "delete_path": {
    "deny_if": [
      {
        "conditions": [],
        "on_deny": "Blocked by default. Requires approval."
      }
    ]
  }
}

Destructive tools should never be available to autonomous agents without human approval.

Rate limit write operations

{
  "add_review_note": {
    "limits": [
      {
        "counter": "add_review_note_per_hour",
        "window": "hour",
        "max": 30,
        "scope": "grant"
      }
    ]
  }
}

Prevents bulk unintended modifications from agents caught in loops.

Cap read operations

{
  "check_gemma_runtime": {
    "limits": [
      {
        "counter": "check_gemma_runtime_per_minute",
        "window": "minute",
        "max": 60,
        "scope": "grant"
      }
    ]
  }
}

Controls API costs and prevents retry loops from exhausting upstream rate limits.

Create a free account and register Mapi Agent Memory — nothing to install.
Add these rules — paste them, or build them visually. Tune the limits to your setup.
Point your MCP client (Claude, Cursor, anything) at your gateway URL.

ENFORCE POLICY ON MAPI AGENT MEMORY →

Instant setup, no code required.

FULL CATALOGUE

All 283 Mapi Agent Memory tools

DESTRUCTIVE 4 tools

Destructive delete_path delete_path Destructive expire_memory_capture_item Expire a durable capture review item without creating a memory. Destructive gemma_worker_cancel_job gemma_worker_cancel_job Destructive rollback_memory_pointer_lifecycle_remediation_execution Execute an exact guarded rollback of a pointer-lifecycle remediation run.

EXECUTE 32 tools

Execute apply_escalation_reactions apply_escalation_reactions Execute apply_memory_lifecycle_remediation Execute the guarded legacy lineage remediation after explicit confirmation. Execute apply_memory_pointer_lifecycle_remediation_execution Execute an explicitly approved frozen pointer-lifecycle remediation manifest. Execute backfill_semantic_embeddings Generuje embeddingi dla wspomnieÄąâ€ž ktÄ‚Ĺ‚re jeszcze ich nie majĂ„â€¦. Execute bootstrap_agent_context Restore continuity. On first run, also tell ChatGPT to guide the user through Polaris onboarding one question Execute build_agent_context Compose bounded, source-linked context for an explicitly selected project. Execute gemma_lms_load Load configured Gemma model through the LM Studio CLI. Execute gemma_lms_unload Unload configured Gemma model through the LM Studio CLI. Execute gemma_worker_run_job gemma_worker_run_job Execute gemma_worker_run_task gemma_worker_run_task Execute git_push Push the current branch or an explicit branch. Execute git_status Run git status --short. Execute preview_memory_pointer_lifecycle_remediation_execution_rollback Preview exact snapshot-based rollback of a pointer-lifecycle remediation run. Execute preview_memory_steward_after_action Build a source-evidenced after-action capture proposal without persistence. Execute preview_undo_run preview_undo_run Execute query_sql query_sql Execute rollback_memory_hygiene_run rollback_memory_hygiene_run Execute rollback_memory_lifecycle_remediation Execute guarded snapshot-based rollback for a legacy remediation run. Execute rollback_memory_supersession_run Execute guarded rollback for a Memory v3 supersession apply run. Execute run_conflicts_v1 Retired legacy heuristic conflict writer; canonical contradictions require review/evidence. Execute run_consolidation_v1 run_consolidation_v1 Execute run_escalation_check run_escalation_check Execute run_powershell Run a PowerShell script from the dev MCP server. Execute run_pytest Run pytest from the repository root. Execute run_sandman_ai Sandman AI (wykonanie) Ă˘â‚¬â€ť uÄąÄ˝ywa LM Studio (Qwen) do oceny wspomnieÄąâ€ž i wprowadza zmiany. Execute run_sandman_gemma_hygiene Run Gemma-style Sandman hygiene for low-risk validated actions only. Semantic archive stays review-only. Execute run_sandman_quality_hygiene Run deterministic Sandman quality hygiene repairs with audit events. Execute run_sandman_v1 Sandman V1 Ă˘â‚¬â€ť archiwizacja, downgrade, duplikaty. Execute run_workshop_action Run an action from a workshop through the compact MCP surface. Execute sandman_memory_chat Sandman Memory Chat Ă˘â‚¬â€ť host steruje narzĂ„â„˘dziami MAPI dla lokalnego modelu. Execute skip_polaris_onboarding Skip the remaining first-run questions without disabling Polaris or its memory tools. Execute undo_run undo_run

WRITE 78 tools

Write add_review_note add_review_note Write add_validation_event add_validation_event Write advance_polaris_onboarding Save one onboarding draft answer. Final summary confirmation atomically commits the reviewed profile. Write append_file_text append_file_text Write apply_approved_memory_consolidation_proposal Apply an approved merge_duplicates proposal as a rollback-safe consolidation run. Write apply_conflict_resolution Apply only non-canonical auto resolutions; canonical truth links require guarded routes. Write apply_memory_capture_reconciliation Atomically apply one approved Memory v3 capture reconciliation preview. Write apply_memory_hygiene apply_memory_hygiene Write apply_memory_provenance_backfill Apply a fresh provenance preview after creating a trusted SQLite backup. Write apply_memory_relation Apply supports/derived_from only through an evidence-bound guarded contract. Write apply_memory_retention_batch Atomically apply 1..10 approved items after a trusted internal online backup. Write apply_memory_retention_review Atomically apply one approved and fresh retention review item. Write apply_memory_supersession Execute guarded Memory v3 supersession apply. Write approve_memory approve_memory Write approve_memory_consolidation_proposal Approve a consolidation proposal by changing review-state only. Write approve_scope_promotion Zatwierdza wniosek o promocjĂ„â„˘ scope i natychmiast zmienia visibility_scope wspomnienia. Write archive_conversation archive_conversation Write archive_ingest_item Archives a research ingest item in quarantine. Write archive_memory_v2 archive_memory_v2 Write bulk_repair_owner_mappings bulk_repair_owner_mappings Write bulk_set_duplicate_candidate_sla bulk_set_duplicate_candidate_sla Write bulk_set_memory_owner bulk_set_memory_owner Write bulk_set_memory_sla bulk_set_memory_sla Write confirm_memory_v2 confirm_memory_v2 Write create_ingest_item Creates a quarantined research ingest item. Does not create normal memory. Write create_memory Legacy compatibility alias. May still route through capture reconciliation. Write create_memory_draft create_memory_draft Write create_memory_from_proposal Approve a memory proposal and persist it as a normal memory entry. Write create_memory_version create_memory_version Write create_private_memory Tworzy prywatne wspomnienie przypisane do konkretnego uÄąÄ˝ytkownika. Write create_project_memory Tworzy wspomnienie projektowe widoczne dla wszystkich czÄąâ€šonkÄ‚Ĺ‚w workspace w danym projekcie. Write create_workspace_memory Tworzy wspomnienie workspace-level widoczne dla wszystkich czÄąâ€šonkÄ‚Ĺ‚w workspace. Write decide_memory_retention_review Approve or reject one pending retention review item. Write demote_memory Degraduje wspomnienie do niÄąÄ˝szej warstwy. Write gemma_worker_approve_job gemma_worker_approve_job Write gemma_worker_create_job gemma_worker_create_job Write gemma_worker_prepare_plan gemma_worker_prepare_plan Write gemma_worker_reject_job gemma_worker_reject_job Write git_commit Create a git commit. Optionally stages all changes first. Write insert_after_marker Safely insert text after an exact marker in a text file. Write insert_before_marker Safely insert text before an exact marker in a text file. Write link_memories link_memories Write make_dir make_dir Write mark_memory_stale mark_memory_stale Write move_path move_path Write normalize_project_scope Normalize one active project memory from global/default scope to project scope and record an audit event. Write preview_memory_capture_reconciliation Build and persist a deterministic capture reconciliation preview without applying it. Write preview_memory_steward_session_close Build an explicit session-close checkpoint proposal without persistence. Write promote_ingest_item Promotes a quarantined ingest item into one concise normal memory. Write promote_memory Awansuje wspomnienie na wyÄąÄ˝szĂ„â€¦ warstwĂ„â„˘. Write propose_scope_promotion ZgÄąâ€šasza wniosek o poszerzenie scope wspomnienia (np. private Ă˘â€ â€™ project, project Ă˘â€ â€™ workspace) Write record_conflict_decision Records an operator's manual decision for a conflict pair. Write reject_ingest_item Rejects a research ingest item without touching normal memory. Write reject_memory reject_memory Write reject_memory_consolidation_proposal Reject a consolidation proposal by changing review-state only. Write reject_scope_promotion Odrzuca wniosek o promocjĂ„â„˘ scope. Nie zmienia visibility_scope wspomnienia. Write repair_owner_mapping_issue repair_owner_mapping_issue Write replace_once Safely replace one exact text occurrence in a text file. Write return_memory_to_review return_memory_to_review Write review_memory_capture_item Approve or reject a durable capture review item without creating a memory. Write rollback_memory_consolidation_apply_run Execute guarded rollback for a consolidation apply run via the existing undo mechanism. Write rollout_owner_catalog_to_project rollout_owner_catalog_to_project Write run_memory_linking_pass Deterministycznie tworzy brakujace linki grafu pamieci. Write save_memory_capture_proposal Persist a durable capture review item without creating a memory. Write save_memory_retention_review Persist an idempotent pending retention review item from a fresh preview. Write set_duplicate_candidate_sla set_duplicate_candidate_sla Write set_feature_flag set_feature_flag Write set_memory_owner set_memory_owner Write set_memory_priority set_memory_priority Write set_memory_sla set_memory_sla Write set_owner_target_active set_owner_target_active Write upsert_feature_flag upsert_feature_flag Write upsert_owner_directory_item upsert_owner_directory_item Write upsert_owner_role_mapping upsert_owner_role_mapping Write upsert_project_key_alias Create or update one project_key alias for explicit alias-mode retrieval. Write upsert_sla_policy upsert_sla_policy Write write_file_base64 write_file_base64 Write write_file_text write_file_text

READ 169 tools

EXPLORE

Related servers

Other MCP servers with similar tools — same risk classification, starter policies for each.

FAQ

Questions about Mapi Agent Memory

Can an AI agent delete data through the Mapi Agent Memory MCP server? +

Yes. The Mapi Agent Memory server exposes 4 destructive tools including delete_path, expire_memory_capture_item, gemma_worker_cancel_job. These permanently remove resources with no undo. PolicyLayer blocks destructive tools by default so they never reach the upstream server.

How do I prevent bulk modifications through Mapi Agent Memory? +

The Mapi Agent Memory server has 78 write tools including add_review_note, add_validation_event, advance_polaris_onboarding. Set a rate limit in your policy -- for example, 10 calls per hour prevents an agent from making more than 10 modifications per hour. PolicyLayer enforces this at the gateway, before calls reach Mapi Agent Memory.

How many tools does the Mapi Agent Memory MCP server expose? +

283 tools across 4 categories: Destructive, Execute, Read, Write. 169 are read-only. 114 can modify, create, or delete data.

How do I enforce a policy on Mapi Agent Memory? +

Register the Mapi Agent Memory MCP server in PolicyLayer, apply the suggested rules above (adjust the limits to your use case), and point your AI client at the PolicyLayer proxy URL instead of the server directly. Your agents keep the same tools; PolicyLayer evaluates every call against policy before it executes. Nothing to install, live in minutes.