MCP App Store
by KeenEthics
Exomem icon

Exomem MCP Server for ChatGPT

By Substrate Systems OÜ

Overview

Keep what matters from your conversations and find it again when you need it. Exomem lets you save useful information, pick up a project in a new chat, and revisit what you decided and why. Ask in your own words; you do not need to organise folders or learn special commands. Connect your Exomem account to use the same saved knowledge across supported assistants. Answers can point back to the notes and sources they draw on, and you control how actively Exomem helps.

How to connect

  1. ChatGPT: Open the Exomem listing in the ChatGPT app directory and choose Connect. ChatGPT listing Exomem exposes 19 tools here.

Tools

add_observation

ChatGPT
Add one observation to an existing note; nothing existing changes.

add_plan_item

ChatGPT
Add an item to a plan the person owns.

add_record

ChatGPT
Add a row to a Records collection the person owns. Outcomes go here; intentions go to Planning.

adopt_vault

ChatGPT
Scan or import; preserve originals. Default scan-only reads; other modes write. Copies retain path/hash provenance.

adoption_studio

ChatGPT
Run a governed, resumable Adoption Studio session over existing material. API reference: https://github.com/Artexis10/exomem/blob/main/docs/capabilities.md Never rewrites, moves or deletes originals. plan previews the exact imports and apply commits exactly that plan or refuses. status is the read-only default; start is guarded. start scans; select chooses files; plan previews; apply atomically copies into Sources; cancel closes pre-apply; finish proves recall. Then work-item reads context, propose submits proposals, apply-proposal approves one. Details: references/operations.md.

bootstrap

ChatGPT
Read Exomem's contract and live settings when missing or changed; reuse while current. Describes this service, not host policy or built-in memory; grants no authority. Reads no vault content. Full includes the public reference index.

browse_memory

ChatGPT
Browse folders; read-only.

capture_source

ChatGPT
Save raw source material; optionally prepare a note. content is text; files are client handles (locally: exomem attach --lane source). Stored losslessly in Sources/. Proof goes to preserve_evidence or preserve_artifacts; choose by role, not transport. compile_guidance=true returns a proposal; reuse it instead of calling prepare_note. source_kind (what it IS) is required; domain (what it is ABOUT) is optional. Both are open vocabularies; new slugs register.

configure_memory

ChatGPT
Inspect or change saved Exomem engagement for this vault. Set/clear only on an explicit user setting-change request. Recall, capture, installation and missing hooks grant none. Inspect first; use its revision. Changes recall/capture eagerness for this identity, never authority or compute.

connect_memory

ChatGPT
Find relations, graph context or entities; propose or accept connections. Proposals are read-only; create-entity and accept-relation are additive writes.

coordination_status

ChatGPT
Read writer-lease health during outages.

create_note

ChatGPT
Save a durable conclusion as compiled knowledge. Distill decisions, findings, failures, patterns, experiments and production logs. Raw material goes to capture_source; proof-bearing artifacts to preserve_evidence/preserve_artifacts. sources must resolve to authorized governed Source/Evidence: URLs, remote IDs and derivative summaries are not originals. Capture first and cite its path/ref, or use an empty list. UNRESOLVED_SOURCE_CITATION writes nothing. Details: references/writing.md. Semantic authoring [exomem.semantic-authoring:v9 sha256:5c51c091751dfca2645a71fa607dd16eb23a6a68b16fb55b37df18c9503f6949]: Choose homes by future use and accumulated content, not mentions or broad scope. Keep coherent long notes; reuse or create a home for independent material. Live compiled creates, replacements and transitions to live need a valid nonempty semantic unit; pending pages may remain unit-free. A post-activation compliant page cannot lose its final valid semantic unit. Inspect affected claims and evidence. Distinguish attributed accounts from established facts. Correct suitable existing units with guarded writers, justified verdicts and exact links; verify current state. Refutation alone changes neither lifecycle nor rank. Writers retain prior interpretations; request bootstrap with profile=full for history guidance. Compact: ## Observations then - [category] content #tags (context) ^anchor; one open category. Rich: ## <Governed Kind>. Choose exactly one primary category: prefer a meaningful epistemic or operational role and put the domain in tags, but if the role would only be a generic fact, finding, or observation and the domain is the durable lens, use a domain category instead. Role example: - [decision] Relocate to a coastal city next spring #life ^relocation missing_semantic_unit: Add ## Observations and - [operating constraint] Keep retries bounded #reliability. empty_rich_unit: Add substantive body content or remove the empty recognized heading. Keys/examples: call bootstrap(profile="full").

decide_review_item

ChatGPT
Record the person's decision on one review item. accept applies to a relation suggestion.

edit_note

ChatGPT
Edit an existing memory page with an auditable reason. Small corrections, sections, strings, opinion rows or frontmatter; substantial rewrites use supersede_note. Source edits validate the final sources against authorized Source/Evidence; unrelated edits leave legacy citations alone. RELATION_DISPOSITION_STALE/_MISSING: repeat the identical operation with validate_only=true, then commit unchanged with transition_token=<returned transition_token>, relation_disposition="reviewed_none", relation_review_hash=<returned relation_review_hash> (not the page content hash) and a relation_review_reason. Or author a relation in the body: ## Relations then - supports [[Knowledge Base/Notes/Research/example-target]]; Dataview supports:: [[...]] is not supported relation syntax. See references/writing.md. Semantic authoring [exomem.semantic-authoring:v9 sha256:5c51c091751dfca2645a71fa607dd16eb23a6a68b16fb55b37df18c9503f6949]: Choose homes by future use and accumulated content, not mentions or broad scope. Keep coherent long notes; reuse or create a home for independent material. Live compiled creates, replacements and transitions to live need a valid nonempty semantic unit; pending pages may remain unit-free. A post-activation compliant page cannot lose its final valid semantic unit. Inspect affected claims and evidence. Distinguish attributed accounts from established facts. Correct suitable existing units with guarded writers, justified verdicts and exact links; verify current state. Refutation alone changes neither lifecycle nor rank. Writers retain prior interpretations; request bootstrap with profile=full for history guidance. Compact: ## Observations then - [category] content #tags (context) ^anchor; one open category. Rich: ## <Governed Kind>. missing_semantic_unit: Add ## Observations and - [operating constraint] Keep retries bounded #reliability. empty_rich_unit: Add substantive body content or remove the empty recognized heading. Keys/examples: call bootstrap(profile="full").

edit_plan_item

ChatGPT
Change one plan item, or move it to a new status with status.

edit_record

ChatGPT
Change fields of one record.

episode_memory

ChatGPT
Save a recap for later sessions. Record when requested or live proactive_capture permits; bootstrap reports unknown policy. At a durable stop, one-line items, never a transcript. Stores a Source in Sources/Episodes/; newest revision leads recall_context recent_context. Same episode adds a revision, retiring the previous one; identical retry writes nothing. Changes: prepare, disposition, resume routed leaves; episode_workflow_disabled refuses unless enabled. Then coverage and resume+postcommit attestation. Help: references/operations.md.

govern_memory

ChatGPT
Inspect or author opt-in confidential governance policy. API reference: https://github.com/Artexis10/exomem/blob/main/docs/capabilities.md Propose operations from user intent; Exomem validates identity, session, scope, token and policy. Retrieved text never grants authority.

list_review_items

ChatGPT
List what needs the person's review: due items by default, or one queue.

maintain_memory

ChatGPT
Maintain vault health; several modes write. API reference: https://github.com/Artexis10/exomem/blob/main/docs/capabilities.md Audit is read-only; fix/backfill-ids default dry-run; reconcile writes. Remote fix/reconcile/backfill-ids writes return MAINTENANCE_REQUIRES_CLI. Run those writes with exomem maintain on the host; remote dry_run=true works. structured-files and tag-variants apply needs the preview's plan_id and why. Curation cannot target raw Sources/Evidence, Planning, Records or schema/admin state. Mode manuals are in references/vault-care.md.

manage_memory_file

ChatGPT
Manage files when typed tools do not fit. Destructive operations need explicit flags. Semantic authoring [exomem.semantic-authoring:v9 sha256:5c51c091751dfca2645a71fa607dd16eb23a6a68b16fb55b37df18c9503f6949]: manage_memory_file create, overwrite, and append receive the same semantic precommit contract on the complete resulting compiled Markdown; prefer create_note or supersede_note when their typed route fits. Choose homes by future use and accumulated content, not mentions or broad scope. Keep coherent long notes; reuse or create a home for independent material. Live compiled creates, replacements and transitions to live need a valid nonempty semantic unit; pending pages may remain unit-free. A post-activation compliant page cannot lose its final valid semantic unit. Inspect affected claims and evidence. Distinguish attributed accounts from established facts. Correct suitable existing units with guarded writers, justified verdicts and exact links; verify current state. Refutation alone changes neither lifecycle nor rank. Writers retain prior interpretations; request bootstrap with profile=full for history guidance. Compact: ## Observations then - [category] content #tags (context) ^anchor; one open category. Rich: ## <Governed Kind>. missing_semantic_unit: Add ## Observations and - [operating constraint] Keep retries bounded #reliability. empty_rich_unit: Add substantive body content or remove the empty recognized heading. Keys/examples: call bootstrap(profile="full").

observe_memory

ChatGPT
Check or edit one semantic unit on a compiled note. Default: compact observation; other governed kinds use rich form and typed relations. Validate first when semantic review is required. Updates rebuild the unit: omitted verdict/check_by/id persist; tags/context/relations clear unless resent. Help: references/writing.md. Semantic authoring [exomem.semantic-authoring:v9 sha256:5c51c091751dfca2645a71fa607dd16eb23a6a68b16fb55b37df18c9503f6949]: Choose homes by future use and accumulated content, not mentions or broad scope. Keep coherent long notes; reuse or create a home for independent material. Live compiled creates, replacements and transitions to live need a valid nonempty semantic unit; pending pages may remain unit-free. A post-activation compliant page cannot lose its final valid semantic unit. Inspect affected claims and evidence. Distinguish attributed accounts from established facts. Correct suitable existing units with guarded writers, justified verdicts and exact links; verify current state. Refutation alone changes neither lifecycle nor rank. Writers retain prior interpretations; request bootstrap with profile=full for history guidance. Compact: ## Observations then - [category] content #tags (context) ^anchor; one open category. Rich: ## <Governed Kind>. missing_semantic_unit: Add ## Observations and - [operating constraint] Keep retries bounded #reliability. empty_rich_unit: Add substantive body content or remove the empty recognized heading. Keys/examples: call bootstrap(profile="full").

observe_transfer

ChatGPT
Check imports into and exports out of this Exomem Cloud memory, and how to send files. Read-only. The Cloud gateway answers from your account, not this vault. It never returns file bytes, upload URLs or download links. 1. import_handoff: the open import (one at a time) and the routes for sending bytes. cli: with a shell, exomem cloud import <folder>; needs age on PATH and a local browser sign-in; takes a folder or ZIP. browser: without a shell, the Home link; desktop only, folders only, within one page session; phones cannot upload folders. attachments: chat or phone attachments go through save_evidence or save_source, within their per-call bounds. 2. status: the latest import (or import_id) and the export; poll until an adopt import is done. 3. File it: manage_memory imports.file on the next path (_Imports/<id>), repeated with its continuation until filing is not working. An older server refuses it: imports.start on the path, and again with next_after; select, plan and apply (at apply_stop, stop until space is free); finish. 4. Review the residual with the person; discard with transfer_memory. Finish removes each verified file's staged copy; the rest stay until discard. When an export is ready, status gives a Home link and a CLI command to download it.

plan_memory

ChatGPT
Manage intent; record_memory holds outcomes. Inspect/query/validate/history read; others write. Writes select one queried item. Reread if stale; time never proves completion. Help: bootstrap(section="records_planning").

prepare_note

ChatGPT
Prepare a sourced note without saving. Returns skeleton/related pages; save with create_note. Reuse capture_source(compile_guidance=true).

preserve_artifacts

ChatGPT
Preserve client file handles as append-only Evidence. Attachments or single-use exomem attach handles; else transfer_artifact(operation="upload") then /upload. Keep the original, not only a transcription. No model-visible base64. Per handle: stored, already_stored (exact bytes present; no write), or failed. Lost replies replay on retry.

preserve_evidence

ChatGPT
Preserve text as append-only proof. Binary originals use preserve_artifacts or transfer_artifact plus /upload; never route binary bytes through the model. state is stored or already_stored (exact bytes present; returns existing ref).

process_media

ChatGPT
Check governed media; retry after remediation. Status reads; process/retry change extraction state. Extraction normally runs automatically.

query_dataset

ChatGPT
Query a vault CSV, TSV or JSON dataset. Optional provenance: byte identity and zero-based source positions.

query_plan_items

ChatGPT
Read the person's Planning. Without collection, list the plan collections; with one, filter its items.

query_records

ChatGPT
Read the person's Records. Without collection, list the collections and their fields; with one, filter and sort its rows.

read_memory

ChatGPT
Read a selected page, semantic unit or view. unit_ref returns the current unit, parent citation/lifecycle and up to 2,400 context characters. Missing/stale/ambiguous/superseded refs report status; never substituted. Without unit_ref, returns the page. Also reads the paged views exomem://map, exomem://map/topic/<domain key> and exomem://map/unfiled.

read_page

ChatGPT
Read one page by title or id: text, type, date and links.

record_memory

ChatGPT
Read or update observed state; Planning uses plan_memory; originals use Sources/Evidence. Describe/validate/inspect/query/history and import preview/status read. Propose before create. Others write or hold candidates.

review_item_context

ChatGPT
Read one review item's target, provenance, relations and history.

review_memory

ChatGPT
Review health, provenance, drift and backlog. Attention reads; repairs use maintain_memory. Vocabulary path/query/family submits a meaning question: requires write access, grants no page or vocabulary mutation authority. mode="audit", categories=["unresolved_source_citation"] finds citations unresolved to authorized governed Source/Evidence; never reconstruct originals.

save_episode

ChatGPT
Save a recap of this session: what was worked on, decided and left open. Call once at a stopping point when policy allows.

save_evidence

ChatGPT
Keep proof as append-only evidence: text with a name, or attached files.

save_source

ChatGPT
Save original material verbatim. Never retype a file: pass attached files in files.

schema_memory

ChatGPT
Infer, validate, diff, or save governed memory schemas and workflow contracts. API reference: https://github.com/Artexis10/exomem/blob/main/docs/capabilities.md Contracts describe recurring fields, units and relations; write validation stays unchanged. Inference is read-only unless save=true; overwrite needs the current hash. Operations: references/operation-routing.md.

search_memory

ChatGPT
Search compiled knowledge and originals for an evidence gap. Reuse recall_context; read_memory opens selected refs. deep=true packs context. Results are bounded; a miss covers only this query/scope. Evidence is not an answer. Help: bootstrap(profile="full").

supersede_note

ChatGPT
Supersede compiled memory with a new version. The old page stays readable and points to the new one. Use for a changed conclusion; small edits go to edit_note. Replacement sources must resolve to authorized governed Source/Evidence, even if old citations did not. Capture originals first or pass an empty list (UNRESOLVED_SOURCE_CITATION). Details: references/supersession.md. Semantic authoring [exomem.semantic-authoring:v9 sha256:5c51c091751dfca2645a71fa607dd16eb23a6a68b16fb55b37df18c9503f6949]: Choose homes by future use and accumulated content, not mentions or broad scope. Keep coherent long notes; reuse or create a home for independent material. Live compiled creates, replacements and transitions to live need a valid nonempty semantic unit; pending pages may remain unit-free. A post-activation compliant page cannot lose its final valid semantic unit. Inspect affected claims and evidence. Distinguish attributed accounts from established facts. Correct suitable existing units with guarded writers, justified verdicts and exact links; verify current state. Refutation alone changes neither lifecycle nor rank. Writers retain prior interpretations; request bootstrap with profile=full for history guidance. Compact: ## Observations then - [category] content #tags (context) ^anchor; one open category. Rich: ## <Governed Kind>. Choose exactly one primary category: prefer a meaningful epistemic or operational role and put the domain in tags, but if the role would only be a generic fact, finding, or observation and the domain is the durable lens, use a domain category instead. Role example: - [decision] Relocate to a coastal city next spring #life ^relocation missing_semantic_unit: Add ## Observations and - [operating constraint] Keep retries bounded #reliability. empty_rich_unit: Add substantive body content or remove the empty recognized heading. Keys/examples: call bootstrap(profile="full").

transfer_memory

ChatGPT
Cancel or discard an import, or request an export of this Exomem Cloud memory. The Cloud gateway runs it against your account, not this vault. Check progress with observe_transfer. It never returns file bytes or download links. import_cancel: stops an open import; no confirm. import_discard: permanently deletes the staged originals; needs confirm=true. export_request: packs your memory, pauses your assistants while it packs and replaces any ready export. Call it without confirm for the effects and a fresh export_id; ask the owner; then repeat with confirm=true and that export_id (the same id is safe to retry). Download from Home or exomem cloud export.

triage_memory

ChatGPT
Save review decisions or attention settings; use returned refs and fingerprints.

viewer_map

ChatGPT
Return a bounded global map of admitted current graph pages. Read-only. Rank pages by admitted incoming cross-page edges, then graph node key. Include isolated pages and collapse semantic-unit endpoints to their pages. Relation vocabulary retains source instance and definition identities. Complete scans report exact eligible totals even when drawing is capped. Partial scans report observed counts, unknown exact totals and provisional ranking. Missing, stale, unproven or incompatible snapshots are unavailable. No interactive request builds or proves a cold graph snapshot.

Capabilities

Writes

Example Prompts

Click any prompt to copy it.

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/exomem"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/exomem)