@prmichaelsen/remember-mcp 3.0.0 → 3.13.0

package/CHANGELOG.md

@@ -5,6 +5,364 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [3.13.0] - 2026-02-28
+
+### Changed
+
+**Ghost Mode — Server-Side Resolution (M16 hardening)**
+
+- Ghost context now resolved server-side via `AuthContext.ghostMode` instead of LLM-accessible `ghost_context` tool parameter
+- `ServerOptions.ghostMode` added to server factory — agentbase.me passes ghost owner/accessor via mcp-auth extras, trust level resolved from Firestore at server creation
+- `GhostModeContext` type added to `src/types/auth.ts` with `owner_user_id`, `accessor_user_id`, `accessor_trust_level`
+- `AuthContext` extended with optional `ghostMode` field
+
+### Removed
+
+- `ghost_context` parameter from `remember_search_memory` and `remember_query_memory` tool schemas — trust levels are no longer LLM-controllable
+- `GhostContext` interface from `search-memory.ts` — replaced by `GhostModeContext` in `types/auth.ts`
+
+### Security
+
+- Trust level escalation prevention: accessor trust is resolved from GhostConfig (Firestore) at server creation time, not from tool arguments
+
+## [3.12.0] - 2026-02-27
+
+### Added
+
+**Ghost System (M16)**
+
+- `remember_ghost_config` tool — manage ghost/persona settings (enable/disable, trust levels, block/unblock users)
+- `src/services/ghost-config.service.ts` — GhostConfig Firestore CRUD (get, set, setUserTrust, blockUser, etc.)
+- `FirestoreGhostConfigProvider` — production GhostConfigProvider implementation
+- `src/services/escalation.service.ts` — `FirestoreEscalationStore` for persistent trust escalation tracking
+- Ghost memory filtering — `content_type: 'ghost'` excluded from default searches (search-memory, query-memory, find-similar, search-space, query-space)
+- `ghost_context` parameter on `remember_search_memory` and `remember_query_memory` — enables trust-filtered cross-user searches
+- `buildTrustFilter` wired into search/query tools when ghost_context present
+- 21 MCP tools total (was 20)
+- 57 new tests (454 total: 453 passed, 1 skipped)
+
+## [3.11.0] - 2026-02-27
+
+### Added
+
+**Trust & Permissions System (M7)**
+
+- `src/types/access-result.ts` — `AccessResult` discriminated union with 6 variants: `granted`, `insufficient_trust`, `blocked`, `no_permission`, `not_found`, `deleted`
+- `src/types/ghost-config.ts` — `GhostConfig` interface (Firestore schema for ghost/persona system), `TrustEnforcementMode` type, `DEFAULT_GHOST_CONFIG` constant
+- `ghost` and `comment` content types added to `ContentType` union and `CONTENT_TYPES` metadata
+- `src/services/trust-enforcement.ts` — 3-mode trust enforcement (query/prompt/hybrid), `buildTrustFilter` for Weaviate, `formatMemoryForPrompt` for 5 trust tiers, trust utilities
+- `src/services/trust-validator.ts` — `validateTrustAssignment` and `suggestTrustLevel` for content-aware trust suggestions
+- `src/services/access-control.ts` — `checkMemoryAccess` with discriminated union results, trust escalation (-0.1 penalty/attempt, block after 3), `resolveAccessorTrustLevel` from GhostConfig, `formatAccessResultMessage`
+- Dependency injection: `GhostConfigProvider` and `EscalationStore` interfaces with in-memory stubs (Firestore in M16)
+- 89 new tests (389 total: 388 passed, 1 skipped)
+
+**Comprehensive Tool Debugging (M8 Task 69)**
+
+- Debug logging integrated into all 18 MCP tools + Weaviate client
+- Pattern: `debug.info` at entry, `debug.trace` for args, `debug.error` in catch blocks
+- 13 remaining tools integrated: create-memory, search-memory, delete-memory, update-memory, find-similar, query-memory, deny, create-relationship, update-relationship, search-relationship, delete-relationship, set-preference, get-preferences
+- 394 tests (393 passed, 1 skipped)
+
+## [3.10.0] - 2026-02-27
+
+### Added
+
+**Moderation System & Per-Space Configuration (M15)**
+
+- `remember_moderate` tool — approve, reject, or remove published memories (requires `can_moderate` permission)
+- `moderation_status`, `moderated_by`, `moderated_at` schema fields on published memory collections
+- `can_moderate` permission added to `GroupPermissions` type
+- Per-space/group configuration via Firestore (`SpaceConfig` service with `require_moderation` and `default_write_mode`)
+- Publish flow sets `moderation_status` based on space/group `require_moderation` config
+- `moderation_filter` parameter on `remember_search_space` and `remember_query_space` tools (default: `approved`)
+- `canModerate` and `canModerateAny` auth helper utilities
+- 20 MCP tools total (was 19)
+- 32 new tests (281 total: 280 passed, 1 skipped)
+
+## [3.9.0] - 2026-02-27
+
+### Added
+
+**Auth Context & Credentials Provider Foundations**
+
+- `src/types/auth.ts` — `AuthContext`, `UserCredentials`, `GroupMembership`, `GroupPermissions`, `CredentialsProvider`, `WriteMode` types
+- `src/services/credentials-provider.ts` — `StubCredentialsProvider` with `createCredentialsProvider()` factory and singleton export
+- 4 ACL schema fields on `PUBLISHED_MEMORY_PROPERTIES`: `write_mode`, `owner_id`, `overwrite_allowed_ids`, `last_revised_by`
+- `authContext?: AuthContext` parameter threaded through all 19 tool handler signatures
+- `server-factory.ts` resolves credentials per request and passes `AuthContext` to all handlers
+- `server.ts` passes null `AuthContext` in standalone mode
+- 3 new tests for `StubCredentialsProvider` (236 total: 235 passed, 1 skipped)
+
+## [3.8.0] - 2026-02-27
+
+### Changed
+
+**V1 → V2 Collection Migration & Code Switch**
+
+- Collection naming switched to v2 dot-notation: `Memory_{SanitizedUserId}` → `Memory_users_{userId}`, `Memory_public` → `Memory_spaces_public`
+- Drop `sanitizeUserId()` — collection names now use literal userId
+- Property renames across all tool files: `type` → `content_type`, `trust` → `trust_score`, `relationships` → `relationship_ids`, `memory_ids` → `related_memory_ids`, `location_gps_lat/lng` → `location_lat/lon`
+- `createMemoryCollection()` and `ensurePublicCollection()` now use v2 schema from `v2-collections.ts`
+- `ALL_MEMORY_PROPERTIES` updated to include both v1 and v2 property names for backward compatibility
+
+### Added
+
+- `scripts/migrate-v1-to-v2.ts` — CLI migration script with `--dry-run`, `--skip-backup`, `--verify-only`, `--batch-size` flags and YAML state file for resumability
+- Schema reconciliation: 25+ missing v1 properties added to `COMMON_MEMORY_PROPERTIES` in `v2-collections.ts` (superset strategy)
+- Migration convenience scripts in `scripts/package.json`
+
+### Fixed
+
+- `RelevantMemory` interface now includes `content_type` property
+- `search-space.spec.ts` filter expectations updated for v2 property names
+
+---
+
+## [3.7.1] - 2026-02-27
+
+### Fixed
+
+**v2 Schema Weaviate Cloud Compatibility**
+
+- Remove reserved `id` property from `COMMON_MEMORY_PROPERTIES` — Weaviate reserves `id` for UUID primary key
+- Change `vectorizer` to `vectorizers` (plural) in all schema creation functions — required by `client.collections.create()` v3 API
+- Replace `'text' as any` data type casts with `configure.dataType.TEXT` (and corresponding typed variants) throughout schema definitions
+- Update `jest.e2e.config.js` to ESM format matching unit test config
+
+### Added
+
+- `src/v2-smoke.e2e.ts` — E2E smoke test validating full v2 flow against live Weaviate (connect → create collections → publish → search → revise)
+
+---
+
+## [3.7.0] - 2026-02-27
+
+### Changed
+
+**remember_revise: Added Confirmation Flow**
+
+`remember_revise` now uses the same two-phase confirmation flow as `remember_publish` and `remember_retract`.
+
+- **Phase 1**: `remember_revise({ memory_id })` validates the request and returns a confirmation token
+- **Phase 2**: `remember_confirm({ token })` executes the revision, syncing content to all published copies
+- Revision logic moved from `revise.ts` to `executeReviseMemory()` in `confirm.ts`
+- New action type: `revise_memory` handled by `remember_confirm`
+- Token response includes destination summary (`space_ids`, `group_ids`, `total_locations`)
+- All existing revision behavior preserved (revision_history, revised_at, partial success)
+
+### Added
+
+**Documentation (Task 173)**
+
+- `agent/design/local.v2-api-reference.md` — Complete API docs for all 6 v2 tools
+- `agent/design/local.v2-migration-guide.md` — v1 → v2 migration guide
+- `agent/design/local.v2-usage-examples.md` — Real-world usage patterns and examples
+- Updated `README.md` with v2 architecture, 18 tools, v2 tool listing
+- Updated `agent/design/local.memory-collection-pattern-v2.md` status to Implemented
+
+---
+
+## [3.6.0] - 2026-02-27
+
+### Added
+
+**Memory Collection Pattern v2 - Tracking Array Initialization (Task 170)**
+
+New memories now initialize `space_ids` and `group_ids` arrays, enabling publication tracking from the moment of creation.
+
+- **`space_ids: []`** initialized on every new memory (managed by `remember_publish` / `remember_retract`)
+- **`group_ids: []`** initialized on every new memory (managed by `remember_publish` / `remember_retract`)
+- **`Memory` interface**: Added `space_ids?: string[]` and `group_ids?: string[]` fields to `src/types/memory.ts`
+- **`remember_update_memory`**: Tracking arrays preserved via spread pattern; explicitly excluded from `UpdateMemoryArgs` to prevent direct modification
+- **Unit Tests** ([`src/tools/create-memory.spec.ts`](src/tools/create-memory.spec.ts))
+  - 19 new tests: `createMemoryTool` schema (8) + `updateMemoryTool` schema (11)
+  - Asserts `space_ids`/`group_ids` are not user-settable via either tool's input schema
+  - Total: 198 tests passing (197 passing, 1 skipped)
+
+## [3.5.0] - 2026-02-27
+
+### Changed
+
+**Memory Collection Pattern v2 - remember_search_space Updated (Task 169)**
+
+`remember_search_space` now targets the v2 collection architecture, adds group search support, and introduces `search_type` selection.
+
+- **Collection target**: `Memory_public` → `Memory_spaces_public` (via `getCollectionName(CollectionType.SPACES)`)
+- **Field filter**: `spaces` property → `space_ids.containsAny()` (matches v2 published memory schema)
+- **`spaces` parameter**: No longer required — omit to search all public memories
+- **`groups` parameter** (new): Array of group IDs — queries `Memory_groups_{groupId}` for each
+- **`search_type` parameter** (new): `"hybrid"` (default) | `"bm25"` (keyword) | `"semantic"` (vector)
+- **All-public search**: When neither `spaces` nor `groups` is provided, queries all of `Memory_spaces_public`
+- **Multi-source merge**: Results from spaces + groups combined, deduplicated by UUID, sorted by score descending
+- **Soft-delete aware**: Added `deleted_at isNull(true)` filter to all queries (requires `indexNullState: true`)
+- **`buildBaseFilters()` helper**: Extracted as exported pure function for testability
+
+### Added
+
+- **Unit Tests** ([`src/tools/search-space.spec.ts`](src/tools/search-space.spec.ts))
+  - 23 new tests: tool schema (9) + `buildBaseFilters` helper (14)
+  - Total: 179 tests passing
+
+---
+
+## [3.4.0] - 2026-02-26
+
+### Added
+
+**Memory Collection Pattern v2 - remember_revise Tool (Task 168)**
+
+Content synchronization for published memories — propagates source memory updates to all published copies.
+
+- **`remember_revise` Tool** ([`src/tools/revise.ts`](src/tools/revise.ts))
+  - Direct action tool (no confirmation flow — content sync is non-destructive)
+  - Reads `space_ids` and `group_ids` tracking arrays from source memory
+  - Updates content in `Memory_spaces_public` (one call for all spaces)
+  - Updates content in each `Memory_groups_{groupId}` in sequence
+  - Partial success: each location reported independently, failures don't block successes
+  - Returns detailed `results` array with status per location
+
+- **Revision History** ([`src/tools/revise.ts`](src/tools/revise.ts))
+  - Old content saved to `revision_history` field (JSON string) before overwrite
+  - Capped at 10 entries (`MAX_REVISION_HISTORY`)
+  - `revised_at` timestamp set on each updated copy
+  - `revision_count` incremented on each updated copy
+  - `parseRevisionHistory()` and `buildRevisionHistory()` exported as pure helpers
+
+- **Server Registration**
+  - `remember_revise` tool registered in both `src/server.ts` and `src/server-factory.ts`
+
+- **Unit Tests** ([`src/tools/revise.spec.ts`](src/tools/revise.spec.ts))
+  - 19 new tests covering `parseRevisionHistory`, `buildRevisionHistory`, and tool definition
+  - Total: 156 tests passing
+
+---
+
+## [3.3.0] - 2026-02-26
+
+### Added
+
+**Memory Collection Pattern v2 - remember_retract Tool (Task 167)**
+
+Selective retraction of published memories from specific spaces and groups, with an orphan strategy that preserves historical data.
+
+- **`remember_retract` Tool** ([`src/tools/retract.ts`](src/tools/retract.ts))
+  - Retract a memory from specific spaces (e.g., `["cooking", "recipes"]`)
+  - Retract a memory from specific groups (e.g., `["group-123"]`)
+  - Selective retraction: retract from some destinations while remaining in others
+  - Ownership validation: users can only retract their own memories
+  - Publication validation: returns error if memory isn't published to the specified destination
+  - Two-phase confirmation flow: generates token, user confirms to execute
+
+- **Orphan Strategy** ([`src/tools/confirm.ts`](src/tools/confirm.ts) `executeRetractMemory()`)
+  - Space memories: remain in `Memory_spaces_public` with `space_ids` array updated (orphaned if empty)
+  - Group memories: remain in `Memory_groups_{groupId}` with `group_ids` array updated (orphaned if empty)
+  - `retracted_at` timestamp set on orphaned memories
+  - Source memory tracking arrays (`space_ids`, `group_ids`) updated on success
+  - Partial success: failed retractions don't block successful ones
+
+- **Server Registration**
+  - `remember_retract` tool registered in both `src/server.ts` and `src/server-factory.ts`
+
+---
+
+## [3.2.0] - 2026-02-26
+
+### Added
+
+**Memory Collection Pattern v2 - remember_publish Tool (Task 166)**
+
+Multi-space and multi-group publication with composite IDs and tracking arrays.
+
+- **Multi-Space Publication** ([`src/tools/publish.ts`](src/tools/publish.ts), [`src/tools/confirm.ts`](src/tools/confirm.ts))
+  - Publish memories to multiple spaces at once (e.g., `["cooking", "recipes"]`)
+  - Publish memories to multiple groups at once (e.g., `["group-123", "group-456"]`)
+  - Dual publication: publish to both spaces and groups simultaneously
+  - Composite IDs (`{userId}.{memoryId}`) for published memories
+  - Tracking arrays (`space_ids`, `group_ids`) on source and published memories
+  - Partial success reporting with detailed results per destination
+  - Error handling for each publication destination independently
+
+- **Tool Schema Updates**
+  - Added `groups` parameter to `remember_publish` tool
+  - Validation for group IDs (no dots allowed)
+  - Support for publishing to spaces only, groups only, or both
+
+### Changed
+
+- `executePublishMemory()` in confirm.ts completely rewritten for v2 architecture
+- Publications now use composite IDs instead of random UUIDs
+- Source memories updated with `space_ids` and `group_ids` tracking arrays
+
+---
+
+## [3.1.0] - 2026-02-26
+
+### Added
+
+**Memory Collection Pattern v2 - Core Infrastructure (Task 165)**
+
+Foundation for multi-user social features including user profiles, friend systems, P2P conversations, and group conversations.
+
+- **Dot Notation Collections** ([`src/collections/dot-notation.ts`](src/collections/dot-notation.ts))
+  - `getCollectionName()` - Generate collection names (Memory_users_{userId}, Memory_spaces_public, Memory_groups_{groupId})
+  - `parseCollectionName()` - Parse collection metadata from names
+  - Collection type validation and ID validation (rejects dots in IDs)
+
+- **Composite IDs** ([`src/collections/composite-ids.ts`](src/collections/composite-ids.ts))
+  - `generateCompositeId()` - Create {userId}.{memoryId} format
+  - `parseCompositeId()` - Extract components from composite IDs
+  - `validateCompositeId()` - Validate and return true on success
+  - `isCompositeId()`, `belongsToUser()` helper functions
+
+- **Tracking Arrays** ([`src/collections/tracking-arrays.ts`](src/collections/tracking-arrays.ts))
+  - Immutable array operations for `space_ids` and `group_ids`
+  - `addToSpaceIds()`, `removeFromSpaceIds()`, `addToGroupIds()`, `removeFromGroupIds()`
+  - Publication status checking (`isPublishedToSpace`, `isPublishedToGroup`, `getPublicationLocations`)
+  - Batch operations for multiple spaces/groups
+
+- **Weaviate Schema Definitions** ([`src/schema/v2-collections.ts`](src/schema/v2-collections.ts))
+  - Schema creation for user, space, and group collections
+  - Property lists and validation functions
+  - Collection name validation and type detection
+
+- **Unit Tests** ([`src/collections/core-infrastructure.spec.ts`](src/collections/core-infrastructure.spec.ts))
+  - 44 tests covering all utilities
+  - 79% coverage on new collections code
+  - 97% coverage on tracking arrays
+
+### Fixed
+
+- Added ID validation to reject dots in user/group IDs in `getCollectionName()`
+- Updated error message in `parseCompositeId()` to include "must be exactly 2 parts"
+- Changed `validateCompositeId()` return type from `void` to `true`
+
+---
+
+## [3.0.1] - 2026-02-25
+
+### Fixed
+
+**CRITICAL: Fixed Weaviate Schema Configuration Bug**
+- Added `indexNullState: true` to inverted index configuration in both schema files
+- Enables filtering on null values (required for `deleted_at IS NULL` queries)
+- Root cause: Soft delete system (v3.0.0) added `deleted_at` field but didn't configure null state indexing
+- Impact: All memory searches were failing with gRPC error "Nullstate must be indexed to be filterable"
+- Modified: [`src/weaviate/schema.ts`](src/weaviate/schema.ts) - Added `invertedIndex` config
+- Modified: [`src/weaviate/space-schema.ts`](src/weaviate/space-schema.ts) - Added `invertedIndex` config
+
+**Migration Required**:
+- ⚠️ Existing Weaviate collections must be recreated to apply this fix
+- Collections created before v3.0.1 will continue to fail on null filtering
+- Delete old collections and restart server to create new collections with correct config
+- Alternative: Export data, delete collections, restart, re-import data
+
+**For Users**:
+- If you encounter "Nullstate must be indexed" error, delete your collections and restart
+- New collections will be created automatically with correct configuration
+- Documents created before `deleted_at` field existed are handled correctly (missing field = null)
+
+---
+
 ## [3.0.0] - 2026-02-25
 
 ### ⚠️ BREAKING CHANGES

package/README.md

@@ -63,7 +63,7 @@ Multi-tenant memory system MCP server with vector search, relationships, and tru
 
 ## Features
 
-- **17 MCP Tools**: Complete CRUD for memories, relationships, preferences, and shared spaces
+- **18 MCP Tools**: Complete CRUD for memories, relationships, preferences, shared spaces, and content sync
 - **Soft Delete with Recovery**: Safe deletion with confirmation flow and recovery capability
 - **Multi-Tenant**: Per-user isolation with secure data boundaries
 - **Shared Spaces**: Publish memories to shared discovery spaces like "The Void"
@@ -150,39 +150,46 @@ await wrapped.start();
 ## Architecture
 
 - **Weaviate**: Vector storage for memories, relationships, and shared spaces
-  - Personal collections: `Memory_{user_id}` (per-user isolation)
-  - Unified public collection: `Memory_public` (all public spaces)
-  - Multi-space support: Memories can belong to multiple spaces via `spaces` array
+  - Personal collections: `Memory_users_{userId}` (per-user isolation)
+  - Public space collection: `Memory_spaces_public` (all shared spaces)
+  - Group collections: `Memory_groups_{groupId}` (private groups)
+  - Composite IDs: `{userId}.{memoryId}` for published memories
+  - Tracking arrays: `space_ids[]` and `group_ids[]` track publication locations
 - **Firestore**: Permissions, preferences, confirmation tokens
   - User data: `users/{user_id}/preferences`, `users/{user_id}/requests`
 - **Firebase Auth**: User authentication
 
-### Multi-Space Architecture (v2.4.0+)
+### Memory Collection Pattern v2 (v3.1.0+)
 
-**Unified Collection**: All public memories stored in single `Memory_public` collection
+Three-tier collection architecture with composite IDs and tracking arrays.
 
-**Benefits**:
-- ✅ Search multiple spaces in one query
-- ✅ Publish to multiple spaces in one operation
-- ✅ No memory duplication
-- ✅ Efficient storage (N× reduction)
+**Collections**:
+- `Memory_users_{userId}` — Private memories with simple IDs
+- `Memory_spaces_public` — All public space memories with composite IDs
+- `Memory_groups_{groupId}` — Group memories with composite IDs
+
+**Key Features**:
+- Publish to multiple spaces and groups simultaneously
+- Composite IDs (`{userId}.{memoryId}`) preserve source reference
+- `remember_revise` syncs content changes to all published copies
+- Orphan strategy keeps retracted memories for historical reference
+- Revision history (max 10 entries) tracks content changes
 
 **Example**:
 ```typescript
-// One memory, three spaces
-{
-  "id": "abc123",
-  "spaces": ["the_void", "dogs", "cats"],
-  "content": "My dog is adorable!",
-  "author_id": "user123"
-}
+// Publish to spaces + groups
+remember_publish({
+  memory_id: "my-recipe",
+  spaces: ["cooking", "recipes"],
+  groups: ["foodie-club"]
+})
 
 // Search across spaces
 remember_search_space({
-  spaces: ["the_void", "dogs"],
-  query: "adorable pets"
+  query: "pasta recipe",
+  spaces: ["cooking"],
+  search_type: "hybrid"
 })
-// Finds memories published to ANY of the requested spaces
 ```
 
 ## Shared Spaces
@@ -195,45 +202,56 @@ Publish memories to shared discovery spaces where other users can find them.
 
 ### Publishing Workflow
 
-1. **Request Publication**: Generate confirmation token
+1. **Publish**: Generate confirmation token
 ```typescript
-// Publish to single space
-remember_publish({ memory_id: "abc123", spaces: ["the_void"] })
-
-// Publish to multiple spaces at once!
-remember_publish({ memory_id: "abc123", spaces: ["the_void", "dogs", "cats"] })
-
+// Publish to spaces + groups
+remember_publish({
+  memory_id: "abc123",
+  spaces: ["the_void", "cooking"],
+  groups: ["foodie-club"]
+})
 // Returns: { success: true, token: "xyz789" }
 ```
 
-2. **User Confirms**: Execute the publication
+2. **Confirm**: Execute the publication
 ```typescript
 remember_confirm({ token: "xyz789" })
-// Returns: {
-//   success: true,
-//   space_memory_id: "new-id",
-//   spaces: ["the_void", "dogs", "cats"]
-// }
+// Creates composite ID copies in Memory_spaces_public and Memory_groups_{groupId}
 ```
 
-3. **Discover**: Search shared spaces
+3. **Revise**: Sync content changes (confirmation required)
 ```typescript
-// Search single space
-remember_search_space({ query: "interesting ideas", spaces: ["the_void"] })
+// After updating source memory, request revision
+remember_revise({ memory_id: "abc123" })
+// Returns: { success: true, token: "xyz789" }
 
-// Search multiple spaces at once!
+remember_confirm({ token: "xyz789" })
+// Updates all copies, preserves old content in revision_history
+```
+
+4. **Retract**: Remove from specific destinations
+```typescript
+remember_retract({ memory_id: "abc123", spaces: ["cooking"] })
+// Orphan strategy: memory remains in collection for historical reference
+```
+
+5. **Search**: Discover shared memories
+```typescript
 remember_search_space({
-  query: "cute dog pictures",
-  spaces: ["the_void", "dogs"]
+  query: "pasta recipe",
+  spaces: ["cooking"],
+  search_type: "hybrid"  // hybrid | bm25 | semantic
 })
 ```
 
-### Space Tools (5 new)
+### Space & Group Tools
 
-- `remember_publish` - Request to publish memory (generates token)
+- `remember_publish` - Publish to spaces and/or groups (confirmation required)
+- `remember_retract` - Retract from spaces and/or groups (confirmation required)
+- `remember_revise` - Sync content to all published copies (confirmation required)
 - `remember_confirm` - Confirm any pending action
 - `remember_deny` - Cancel any pending action
-- `remember_search_space` - Search shared spaces
+- `remember_search_space` - Search shared spaces and groups
 - `remember_query_space` - Ask questions about shared memories
 
 ## Safe Deletion with Confirmation
@@ -366,8 +384,13 @@ REMEMBER_MCP_DEBUG_LEVEL=TRACE npm start
 
 ## Documentation
 
-See `agent/` directory for:
-- Design documents (`agent/design/`)
+See `agent/design/` for detailed documentation:
+- [Memory Collection Pattern v2](agent/design/local.memory-collection-pattern-v2.md) — Architecture and design rationale
+- [v2 API Reference](agent/design/local.v2-api-reference.md) — Complete tool schemas and parameters
+- [v2 Migration Guide](agent/design/local.v2-migration-guide.md) — Migrating from v1 to v2
+- [v2 Usage Examples](agent/design/local.v2-usage-examples.md) — Real-world usage patterns
+
+Additional project docs:
 - Milestones (`agent/milestones/`)
 - Implementation tasks (`agent/tasks/`)