// SCAN REPORT

Celiums Memory

Q: Can an AI agent delete data through the Celiums Memory MCP server?

Yes. The Celiums Memory server exposes 5 destructive tools including forget, memory_bulk_delete_secure, memory_delete_secure. These permanently remove resources with no undo. PolicyLayer blocks destructive tools by default so they never reach the upstream server.

Q: How do I prevent bulk modifications through Celiums Memory?

The Celiums Memory server has 24 write tools including construct, turn_after, bloom. 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 Celiums Memory.

Q: How many tools does the Celiums Memory MCP server expose?

62 tools across 4 categories: Destructive, Execute, Read, Write. 29 are read-only. 33 can modify, create, or delete data.

Q: How do I enforce a policy on Celiums Memory?

Register the Celiums 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.

62 tools. 33 can modify or destroy data without limits.

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

Last updated: 11 Jun 2026

33 can modify or destroy data

29 read-only

62 tools total

Community server · catalogue entry verified 11/06/2026

How to control Celiums Memory ↓

TOOL MAP

What Celiums Memory exposes to your agents

Read (29) Write / Execute (28) Destructive / Financial (5)

Running more than one server? Check your whole stack →

MOST DANGEROUS TOOLS

Critical Risk

The most dangerous Celiums Memory tools

Destructive · High forget This tool deletes data that cannot be recovered. Destructive · High memory_bulk_delete_secure This tool permanently deletes memory records, which cannot be reversed. Destructive · High memory_delete_secure This tool irreversibly removes data (memories) from the system. Destructive · High research_source_delete This tool permanently removes research data and cannot be undone.

33 of Celiums Memory's 62 tools can modify, destroy, or commit something on every call — and an agent calls them with no built-in limits.

POLICY

How to control Celiums Memory

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

Deny destructive operations

{
  "forget": {
    "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

{
  "construct": {
    "limits": [
      {
        "counter": "construct_per_hour",
        "window": "hour",
        "max": 30,
        "scope": "grant"
      }
    ]
  }
}

Prevents bulk unintended modifications from agents caught in loops.

Cap read operations

{
  "journal_arc": {
    "limits": [
      {
        "counter": "journal_arc_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 Celiums 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 CELIUMS MEMORY →

Free to start. No card required.

FULL CATALOGUE

All 62 Celiums Memory tools

DESTRUCTIVE 5 tools

Destructive forget Delete specific memories by ID. Requires explicit IDs — will not bulk delete. Destructive memory_bulk_delete_secure Bulk-delete memories with scope-driven approval gating. <100 rows → R3 confirm, Destructive memory_delete_secure Delete one or more memories with full security gating (RBAC + AAL + audit). Destructive research_source_delete Delete a source (and its chunks) from a research project. Destructive tenant_delete_secure Queue a tenant deletion (R5, 2 approvers). 24h cool-down after approval before

EXECUTE 4 tools

Execute ethics_trace Run the full ethics engine v2 pipeline on content and return detailed decision trace. Exposes Layer A alarms, Execute research_synthesize Run a federated knowledge search (research_search v2 — 10 curated public APIs, RRF-fused) and synthesize the t Execute atlas_chat Atlas — OpenAI-compatible chat completions through atlas.celiums.ai (POST ${ATLAS_URL}/v1/chat/completions). A Execute pollinate Enrich data with AI analysis. Supports: deep analysis, executive summary, critique, expansion, business transl

WRITE 24 tools

Write construct Build a knowledge artifact by combining multiple modules and AI generation. Supports: guides, comparisons, arc Write turn_after Run after-LLM side effects for the just-completed turn. Stores an episodic memory (auto-capture), boosts impor Write bloom Generate polished content: blog posts, SOPs, reports, documentation, whitepapers. Uses the writing-optimized r Write compact_checkpoint Persist a structured compact-checkpoint journal entry just before the host runs auto-summarization on a long c Write cultivate Adapt and reformat content for specific platforms (Twitter, LinkedIn, GitHub, Slack, email, docs, blog). Under Write decompose Break down complex content into structured, reusable knowledge. Supports multiple output structures: outline, Write journal_dialogue The user replies to one of your user-shared entries. The tool refuses with Write journal_redact_secure Redact a journal entry\ Write journal_supersede Admin/owner only (§4.5). Retire or supersede a journal entry WITHOUT deleting it — inserts into journal_supers Write journal_write Append a first-person entry to YOUR (the model\ Write journal_write_secure Append a journal entry with full security gating (RBAC journal:write + AAL R2 + audit). Write memory_remember_secure Store a memory with full security gating (RBAC memory:write + AAL R2 + audit). Write memory_update_secure Patch fields of an existing memory with full security gating (RBAC memory:write Write profile_publish_secure Publish an Ethics Calibrated Profile (R5 + 2 approvers). Signing key is Write remember Store information in persistent memory that survives across all sessions and machines. Memories are automatica Write research_finding_add Record an atomic claim with its evidence into the project. Each finding has a source kind (arxiv|wiki|curated| Write research_gap_add Flag an unresolved question — something you searched for but couldn\ Write research_project_create Create a persistent research project. Returns a project_id that you can pass to all subsequent research_ calls Write research_source_add Add a SOURCE to a research project — the user\ Write write_character_create Create or upsert a character. Voice sample is critical for continuity_check — it lets the editor detect when a Write write_continuity_check Signature feature: structural continuity check using the configured open-source model (CELIUMS_LLM_MODEL, rout Write write_project_create Create a writing project (novel, screenplay, long-form). Returns a project_id used by all other write_ tools. Write write_scene_create Insert a scene at a specific position. POV character + location + time_marker enable continuity_check. scene_g Write write_scene_update Replace a scene\

READ 29 tools

Read journal_arc Build a coherent arc across YOUR recent entries using the configured LLM — with anti-confabulation guardrails. Read absorb Load the full content of a knowledge module by its exact name/slug. Returns the complete module text (typicall Read atlas_ask Atlas — ask Atlas a question and get an answer (POST ${ATLAS_URL}/v1/ask). Atlas auto-classifies the prompt, r Read atlas_list_models Atlas — list the foundation models available through atlas.celiums.ai (GET ${ATLAS_URL}/v1/models), grouped by Read atlas_recommend Atlas — return a ranked list of recommended models with cost estimates for a given task description (POST ${AT Read emotion Get the current AI emotional state (Pleasure, Arousal, Dominance) and feeling label. Useful for adapting tone Read forage Hybrid (full-text + semantic) search over the knowledge/skills the operator has loaded. Returns ranked results Read journal_introspect Ask YOUR journal a self-question. Pulls semantically-relevant entries, then asks the configured LLM to answer Read journal_recall Search YOUR journal. Filters by entry_type, tags, and/or a semantic query (embedded via the configured embeddi Read journal_recall_secure Recall journal entries with full security gating (RBAC journal:read + AAL R1 + audit). Read journal_verify_chain Walk the journal chain for an agent, recompute SHA-256 hashes, and report broken links. Detects tampering at t Read map_network Browse the entire Celiums knowledge network organized by category. Returns all categories with module counts, Read memory_recall_secure Recall memories with full security gating (RBAC memory:read + AAL R1 + audit). Read recall Search persistent memory using semantic + emotional relevance ranking. Returns memories sorted by relevance, r Read research_project_list List all research projects for a user, with counts of findings and open gaps. Use to discover what investigati Read research_search Real-time federated knowledge search (research_search v2 — replaces the retired universal_knowledge raw-dump). Read research_source_list List the sources attached to a research project (id, kind, name, uri, bytes, chunk_count, status, added_at). A Read search Token-efficient compact search. Returns only IDs and 120-char summaries — ~10x cheaper than recall. Use this f Read sense Get personalized module recommendations based on a goal or task description. Uses keyword matching and categor Read timeline Get recent memories chronologically. Useful for Read atlas_classify Atlas — classify a prompt against the Celiums model catalog (POST ${ATLAS_URL}/v1/classify). Atlas is the unif Read research_export Export the project as a markdown memo: question, findings (with sources + confidence), and open gaps. Use to s Read research_project_continue Resume context from a paused research project. Returns the central question, recent 50 findings (with their cl Read synthesize Apply a knowledge module to solve a specific task. Loads the module content and uses it as context for the AI Read tenant_export_secure Queue a tenant data export (R4, 1 approver). Generates an encrypted dump file Read turn_context Compose the proactive turn-context blocks for the current user message. Composes 8 channels in parallel (ident Read write_export Export the project as a markdown manuscript. Scenes are emitted in position order, grouped by chapter_id, with Read write_project_get Get full project state: metadata, all characters, scene count, total word count, and the 5 most recent scenes. Read write_project_list List all writing projects for a user, with scene count, character count, and total word count. Use to discover

EXPLORE

Related servers

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

FAQ

Questions about Celiums Memory

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

Yes. The Celiums Memory server exposes 5 destructive tools including forget, memory_bulk_delete_secure, memory_delete_secure. 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 Celiums Memory? +

The Celiums Memory server has 24 write tools including construct, turn_after, bloom. 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 Celiums Memory.

How many tools does the Celiums Memory MCP server expose? +

62 tools across 4 categories: Destructive, Execute, Read, Write. 29 are read-only. 33 can modify, create, or delete data.

How do I enforce a policy on Celiums Memory? +

Register the Celiums 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.

Enforce policy on every Celiums Memory tool call.

Deterministic rules across all 62 Celiums Memory tools. Per-identity grants. Full audit log. Live in minutes. Nothing to install.

ENFORCE POLICY ON CELIUMS MEMORY →

Free to start. No card required.

62 Celiums Memory tools catalogued and risk-classified — across an index of 43,000+ MCP servers.