Reverie — Watchlist | MCP100 Index

MCP100 Assessment

Watchlist / Tracking

Why we’re tracking it

Follow-up value is specific to Reverie typed knowledge-graph memory served over MCP: Do package renaming and graph consolidation preserve entity provenance when migrating from the legacy memory package?

Reviewed Oct 5, 2026 · Documentation review; complete setup workflow not independently tested

Best fit

Persistent relationship-aware agent memory in Neo4j.

What we know

Self-hosted Neo4j graph-memory MCP with entity/relationship CRUD, hybrid search, traversal, read-only Cypher, statistics and deterministic graph consolidation.

Maintenance / activity evidence

@knowall-ai/reverie 0.5.3 published 2026-09-24T23:56:42.265Z

Tracking evidence observed Oct 5, 2026. These dated signals do not establish current maintenance. Reviewed enrichment confidence: high.

Verification gaps / watch-outs

Installation and comprehensive security have not been independently audited.
Graph memory CRUD and consolidation authority; HTTP Brain API read access when served. Writes: Create/update/delete memories/connections; Consolidate/merge graph; Refresh embeddings.
Neo4j credentials + optional HTTP/embedding credentials.
Sensitive access: Graph memory, Neo4j credentials, Potential private organizational knowledge

Persistent memory may contain secrets; destructive merge/delete actions need dry-run/review.

Full authority, access and evidence limits are available in Review context and Access & risk.

All limitations / Review context

Do package renaming and graph consolidation preserve entity provenance when migrating from the legacy memory package?

The bounded review of the publisher's documented Reverie typed knowledge-graph memory served over MCP and release notices identified no disclosed material condition preventing an accurately scoped public tracking entry. The resource's operational permissions remain uncertainty; this assessment does not certify installation safety.

Persistent memory may contain secrets; destructive merge/delete actions need dry-run/review.
Graph memory CRUD and consolidation authority; HTTP Brain API read access when served.
Neo4j credentials + optional HTTP/embedding credentials.
Primary storage is Neo4j; local embeddings default possible, with optional external providers.
Reviewed write actions: Create/update/delete memories/connections; Consolidate/merge graph; Refresh embeddings
Sensitive access: Graph memory, Neo4j credentials, Potential private organizational knowledge
Installation, client compatibility and comprehensive security have not been independently audited. Authentication and access depend on the configured services.
Runtime behavior, dependencies and undisclosed vulnerabilities were not independently audited.
Documented tool authority and any remote data handling must be checked before installation or use.
Tool inventory, client matrix and permissions completeness were not requirements for this Radar tracking review.
Requires Neo4j 5.9+ and Node 18+.
APOC needed for duplicate merging.
Delete/dream operations alter graph state.
External embedding providers may receive memory text.
Current repo/package is the documented successor.
Preserve successor relationship to the legacy package.
Successor/context: @knowall-ai/reverie succeeds legacy @knowall-ai/mcp-neo4j-agent-memory

Original activity observations

@knowall-ai/reverie 0.5.3 published 2026-09-24T23:56:42.265Z

Technical identity

npm/@knowall-ai/reverie

Exact implementation source: https://www.npmjs.com/package/@knowall-ai/reverie ↗
Reviewed resource type: MCP Server
Repository identity: knowall-ai/mcp-reverie
Package identity: npm:@knowall-ai/reverie
Service endpoint: http://127.0.0.1:8643

Evidence & Sources

Evidence basis: Recorded source evidence
Setup runtime verification: Not independently runtime-tested by MCP100 for this setup workflow.

Reviewed Oct 5, 2026 · Latest source observation Oct 8, 2026

Package registry ↗
@knowall-ai/reverie
Recorded package identity, distribution or download evidence.
Includes references whose attribution is not established.
GitHub repository ↗
knowall-ai/mcp-reverie
Repository reference; implementation attribution is not established for every document.
Includes references whose attribution is not established.

Setup · Watchlist

Install & Agents

Publisher instructions.

Source evidence observed Oct 14, 2026.

Additional setup details required — see the specific gaps below.

Warning

Publisher documentation reviewed; compatibility with user environment and live service not tested.
Disconnecting MCP does not erase local memory store. Back up/delete data only by deliberate user choice.
Documented by publisher; environment and client prerequisites still required. Placeholder values must not be executed verbatim.
MCP100 normalized this JSON from publisher-documented package, env and/or client examples; not verified byte-for-byte against a live client. This is not Codex config.toml or VS Code servers JSON.
First check client tool listing and Neo4j connection using a non-mutating read; do not test by deleting memory.
Never paste credentials into chat.
MCP client unregister, stopping hosted processes, package uninstall, OAuth disconnect and token revocation are distinct actions.

Install / connect
Publisher instructions
Select embedding backend (OpenAI, Azure, Ollama or Voyage) and configure provider credentials where applicable.
publisher install or stdio launch · Publisher instructions
npx -y @knowall-ai/reverie
Configure / authenticate
MCP100 setup guide
Requires Neo4j configuration; npx starts Reverie server, not a complete Neo4j database deployment.
claude desktop or cursor mcpServers json · MCP100 composed guide
{ "mcpServers": { "reverie": { "args": [ "-y", "@knowall-ai/reverie" ], "command": "npx" } } }
Check connection
Suggested check
After connecting Neo4j, list/inspect exposed Reverie tools, then use a read-only graph retrieval; avoid writing/removing memories.
Remove / uninstall / revoke
MCP100 setup guide
If you added this integration to Claude Code under the registration name `reverie`, run `claude mcp remove reverie` to unregister that client entry. In Claude Desktop, Cursor, VS Code or another host, delete/disable the corresponding named MCP registration in that host's settings and restart/reconnect. If you installed an npm package globally or operate a separate HTTP server, uninstall/stop it with its original deployment method. This does not delete user data or revoke service credentials.
MCP client unregister, stopping hosted processes, package uninstall, OAuth disconnect and token revocation are distinct actions.
A claude mcp remove command is only valid for Claude Code when the local registration actually has that name. Other MCP hosts require their own UI/config cleanup.
Separate credential guidance / limitations: Reverie local stdio MCP uses Neo4j user/password and optional embedding vendor keys. Hosted serve has a self-provisioned REVERIE_SERVE_TOKEN; rotate that deployment secret and restart/redeploy consumers, and manage Neo4j/vendor keys with their issuers. No Reverie account to revoke.
Remove client registration · Claude Code only · MCP100 composed guide
claude mcp remove reverie

Official Documentation & Sources

What it does

Self-hosted Neo4j graph-memory MCP with entity/relationship CRUD, hybrid search, traversal, read-only Cypher, statistics and deterministic graph consolidation.

Capability: Files & Knowledge
Resource type: MCP Server
Best fit: Persistent relationship-aware agent memory in Neo4j.

Representative uses

Store facts/entities
Connect people/projects
Hybrid memory search
Traverse related memories
Run read-only Cypher
Dry-run/execute graph consolidation

Limitations

Requires Neo4j 5.9+ and Node 18+.
APOC needed for duplicate merging.
Delete/dream operations alter graph state.
External embedding providers may receive memory text.
Current repo/package is the documented successor.
Preserve successor relationship to the legacy package.
Successor/context: @knowall-ai/reverie succeeds legacy @knowall-ai/mcp-neo4j-agent-memory

Reviewed access

Access & risk

Review the exact access boundary before connecting this MCP to accounts, credentials, or sensitive data.

Authentication: Neo4j credentials + optional HTTP/embedding credentials.
Permissions: Graph memory CRUD and consolidation authority; HTTP Brain API read access when served.
Data handling: Primary storage is Neo4j; local embeddings default possible, with optional external providers.
Cost: self_hosted_plus_provider
Risk and cautions: Persistent memory may contain secrets; destructive merge/delete actions need dry-run/review.

Access scope and cautions

Write actions: Create/update/delete memories/connections Consolidate/merge graph Refresh embeddings
Sensitive access: Graph memory Neo4j credentials Potential private organizational knowledge
Risk level: high

FAQ

Based on reviewed evidence as of 2026-10-08. Watchlist tracking is not a recommendation or a security certification.

Which tasks are documented for this implementation?

The reviewed tasks include: Store facts/entities; Connect people/projects; Hybrid memory search; Traverse related memories; Run read-only Cypher; Dry-run/execute graph consolidation.

What setup and authentication are documented?

Recorded installation/entrypoint step: npx -y @knowall-ai/reverie. Recorded endpoint: http://127.0.0.1:8643. Recorded transport: stdio, streamable_http. Recorded environment-variable notes: NEO4J_URI; NEO4J_USERNAME; NEO4J_PASSWORD; NEO4J_DATABASE; REVERIE_SERVE_TOKEN (HTTP); embedding-provider vars as applicable. No configuration values are supplied here. Prerequisites: Neo4j 5.9+; Node.js 18+; APOC for merging. These are the documented setup details, not a claim that installation or client compatibility was independently tested. Authentication: Neo4j credentials; bearer token for HTTP; optional embedding-provider keys.

Can it modify data or perform write operations?

Documented write actions include: Create/update/delete memories/connections; Consolidate/merge graph; Refresh embeddings. Recorded permission boundary: Graph memory CRUD and consolidation authority; HTTP Brain API read access when served.

What data-handling boundary should I consider?

Primary storage is Neo4j; local embeddings default possible, with optional external providers. Recorded sensitive-access context: Graph memory; Neo4j credentials; Potential private organizational knowledge.

What limitations and risks should I check before use?

Persistent memory may contain secrets; destructive merge/delete actions need dry-run/review. Requires Neo4j 5.9+ and Node 18+. APOC needed for duplicate merging. Delete/dream operations alter graph state. External embedding providers may receive memory text. Current repo/package is the documented successor. Preserve successor relationship to the legacy package. Successor/context: @knowall-ai/reverie succeeds legacy @knowall-ai/mcp-neo4j-agent-memory. Runtime behavior, dependencies and undisclosed vulnerabilities were not independently audited. Documented tool authority and any remote data handling must be checked before installation or use.

Is this an official MCP?

KnowAll AI is listed as First-party MCP in MCP100’s current evidence.