ENTERPRISE_v3.1

Wyrm

Enterprise AI Memory System

Persistent, searchable memory for AI assistants with prompt caching, real-time usage tracking, and cost optimization. Built for enterprise teams who need security, performance, and visibility.

// CAPABILITIES

Enterprise-Grade AI Memory

Persistent Memory

SQLite-backed storage with WAL mode. AI remembers across sessions with zero data loss.

Prompt Caching

In-memory cache with auto-invalidation on writes. Sub-millisecond responses for read-heavy workloads.

Full-Text Search

FTS5 search engine across sessions, quests, and data. Instant results at any scale.

AES-256 Encryption

Military-grade AES-256-GCM encryption for sensitive data at rest.

Usage & Cost Tracking

Real-time token tracking, cache hit rates, and cost estimation. Know exactly what you spend.

MCP Protocol

Native Model Context Protocol with cache_control hints for Anthropic prompt caching.

Data Lake

Batch insert/query operations with category-based organization. No limits on storage.

Version Sync

Bi-directional sync with .wyrm/ markdown files. Git-friendly context management.

// v3.1.0

What's New

Wyrm v3.1 introduces prompt caching for sub-millisecond reads, real-time token usage tracking with cost estimation, and MCP cache_control hints for Anthropic models. Monitor every token, know exactly what you spend.

In-memory MemCache for 8 read-only tools
cache_control MCP hints for Anthropic prompt caching
wyrm_usage tool with token/cost tracking
Auto-cache invalidation on any write operation
Cost estimation at Claude Opus rates ($15/M in, $75/M out)
Response fingerprinting for delta detection
Top tools breakdown by token consumption
500-entry rolling usage log
// TECHNICAL

Built for Production

Sub-ms Cache

MemCache layer returns cached reads in <1ms. Auto-invalidates on any write operation.

Cost Optimization

Track input/output tokens, cache savings, and net cost per session. Budget your AI spend.

Token Tracking

Every tool call tracked — tokens in, tokens out, response time, cache hit/miss.

Quest Management

Task tracking with priorities, tags, and status. Project management built into your AI.

WAL Mode

Write-Ahead Logging for concurrent access. Multiple AI clients, one database, zero conflicts.

Enterprise Security

RBAC-ready architecture, audit logging, encrypted storage. Built for compliance.

Tech Specs
SQLite with WAL mode for concurrent access
FTS5 full-text search engine
AES-256-GCM encryption at rest
MCP protocol with cache_control hints
In-memory cache with auto-invalidation
Real-time token usage tracking
Cost estimation (Claude Opus rates)
TypeScript with full type safety
Zero external runtime dependencies
// LICENSING

Enterprise Plans

Contact sales for pricing and deployment options.

Team

Small teams getting started

Up to 10 users
Unlimited storage
Full-text search
Usage tracking
Email support
POPULAR

Enterprise

Advanced features for scale

Unlimited users
AES-256 encryption
Prompt caching
Cost optimization
Priority support
Custom integrations

Enterprise Plus

High-security environments

On-premise deployment
Custom SLA
Audit logs
RBAC
24/7 support
Dedicated engineer

Your AI, Your Memory

Stop losing context between sessions. Wyrm gives your AI persistent memory with the performance and security your enterprise demands.