Skip to main content
The Velixar MCP server is a cognitive memory layer for AI assistants. It goes beyond store-and-retrieve — your assistant gets orientation, contradiction detection, knowledge graph traversal, identity awareness, and belief tracking that persists across every session. Works with Claude Desktop, Kiro, Cursor, Windsurf, Continue.dev, and any MCP-compatible client.

​
Install

npm install -g velixar-mcp-server

​
Configure

Get an API key from Settings → API Keys. 
{
  "mcpServers": {
    "velixar": {
      "command": "velixar-mcp-server",
      "env": {
        "VELIXAR_API_KEY": "vlx_your_key"
      }
    }
  }
}
Restart your assistant. The 25 tools, resources, and prompts are immediately available.

​
What Makes This Different

Most memory tools give you store and search. Velixar gives your assistant a cognitive layer:

Orient in one call

velixar_context synthesizes a workspace briefing from memories, identity, and graph data — your assistant knows where it is immediately.

Knowledge graph

velixar_graph_traverse walks entity relationships. Your assistant understands how concepts, people, and systems connect.

Contradiction detection

velixar_contradictions surfaces conflicting beliefs with resolution guidance. No more silently holding two incompatible facts.

Belief evolution

velixar_timeline shows how understanding of a topic changed over time — not just what’s true now, but how you got here.

Identity awareness

velixar_identity maintains a persistent model of who you are — preferences, expertise, goals, constraints.

Session continuity

velixar_session_save and velixar_session_recall let your assistant pick up exactly where it left off.

​
Tools

​
Memory (5 tools)

ToolDescription
velixar_storeStore a memory with tags, tier, and type
velixar_searchSemantic search across memories
velixar_listBrowse with pagination
velixar_updateEdit content or tags
velixar_deleteRemove a memory

​
Cognitive (8 tools)

ToolDescription
velixar_contextSynthesized workspace briefing
velixar_identityUser profile — get, store, or update
velixar_contradictionsConflicting facts with resolution guidance
velixar_timelineHow a topic evolved over time
velixar_patternsRecurring problem/solution motifs
velixar_inspectDeep memory inspection with provenance
velixar_graph_traverseWalk entity relationships
velixar_distillExtract durable memories from sessions

​
Lifecycle (8 tools)

ToolDescription
velixar_session_saveSave session for later recall
velixar_session_recallRestore previous session context
velixar_batch_storeStore up to 20 memories at once
velixar_batch_searchRun up to 10 queries simultaneously
velixar_consolidateMerge related memories
velixar_retagBulk update tags
velixar_exportExport as JSON or Markdown
velixar_importBulk import from JSON, Markdown, Notion, Obsidian

​
System (4 tools)

ToolDescription
velixar_healthBackend status and latency
velixar_debugDiagnostics and cache state
velixar_capabilitiesFeature inventory
velixar_securityContent scanning mode

​
Live Resources

Resources inject context into your assistant automatically — no tool call needed.
ResourceContent
velixar://system/constitutionBehavioral rules and cognitive modes
velixar://identity/currentYour persistent user profile
velixar://memories/recentRecent memories (compact)
velixar://memories/relevantContextually relevant memories
velixar://domains/{domain}/shadow_graphKnowledge graph for a domain
Hosts that don’t support resources (like Cursor) still get full tool access. The same information is available via velixar_context, velixar_identity, and velixar_capabilities.

​
Workflow Prompts

16 built-in prompts guide multi-step reasoning:
GroupPrompts
OrientationRecall prior reasoning, build project context, profile entity, orient-then-narrow
ConflictResolve contradiction, identify knowledge gaps
ContinuityTrace belief evolution, resume session, reconstruct decision path
LifecycleDistill session, consolidate topic, retag recent
IdentitySummarize identity, detect preference shift, align response style
EnterpriseEvaluate enterprise fit

​
Workspace Isolation

Memories are scoped per workspace. Projects never bleed into each other.
PrioritySource
1VELIXAR_WORKSPACE_ID env var
2.velixar.json in project root
3Git root directory name (automatic)

​
Environment Variables

VariableRequiredDescription
VELIXAR_API_KEYYesYour API key (vlx_...)
VELIXAR_WORKSPACE_IDNoExplicit workspace scope
VELIXAR_API_URLNoCustom API endpoint
VELIXAR_USER_IDNoUser ID for scoping
VELIXAR_DEBUGNotrue for verbose logging
VELIXAR_LOG_FORMATNojson for structured logging
VELIXAR_HEALTH_PORTNoHTTP health check port

​
Reliability

The server includes automatic retry with exponential backoff, a circuit breaker that opens after sustained failures and auto-recovers, and cache fallback that serves stale data during outages rather than failing.

​
Source

github.com/VelixarAi/velixar-mcp-server