kratos-mcp
kratos-mcp is an MCP server that provides a local “memory” layer for AI coding tools. It auto-detects projects, stores memories in per-project SQLite databases (optionally global/shared), and exposes MCP tools for saving/searching/retrieving/forgetting memories plus basic project management and a local security_scan for detecting PII/secrets in text before saving.
Score Breakdown
⚙ Agent Friendliness
🔒 Security
Security_scan is described as a local pre-save check for PII/secrets, and the README claims data stays local (SQLite) with per-project isolation. However, the provided content does not document how secrets are handled internally (logging/redaction), whether any access control exists for MCP tool invocation, or whether the server enforces safe path handling for storage changes.
⚡ Reliability
Best When
Used locally by a developer alongside an MCP-compatible AI coding agent where preserving context across sessions is valuable and local persistence is acceptable.
Avoid When
Avoid when you need multi-user access control, hosted uptime/SLA, or strong contractual API semantics for external integrations; also avoid if you cannot ensure the local storage path and memory contents are safe to persist.
Use Cases
- • Persisting architecture decisions, coding standards, and reusable implementation notes across AI coding sessions
- • Semantic/project-scoped retrieval of prior knowledge to reduce re-explaining work
- • Maintaining separate memory stores per repo to avoid cross-project contamination
- • Pre-save scanning for potential PII/secrets before storing memory
- • Switching or migrating storage locations for local developer workflows
Not For
- • A hosted/team memory service that needs centralized governance or online reliability guarantees
- • Use cases requiring fine-grained, server-side access control (the README suggests local-only storage with no documented auth)
- • High-concurrency/multi-tenant deployments without additional operational hardening
- • Public APIs for arbitrary external callers (no REST/GraphQL endpoints described)
Interface
Authentication
README describes local MCP server configuration for Claude Desktop/Claude Code/Cursor via launching npx/command. No API keys, OAuth, or scope model is described for tool access.
Pricing
No pricing information provided; npm package implies free open-source usage per MIT license, but runtime infra costs (local compute/disk) are user-controlled.
Agent Metadata
Known Gotchas
- ⚠ No documented auth or access boundaries; agents with MCP connectivity can potentially invoke memory operations on the host.
- ⚠ Storage is local and stateful; changes in storage path/project switching may affect which memory database is used.
- ⚠ No documented rate limits, concurrency limits, or backoff/retry strategy for tool calls.
Alternatives
Full Evaluation Report
Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for kratos-mcp.
AI-powered analysis · PDF + markdown · Delivered within 30 minutes
Package Brief
Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.
Delivered within 10 minutes
Score Monitoring
Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.
Continuous monitoring
Scores are editorial opinions as of 2026-03-30.