Ratine
https://github.com/goweft/ratine
Only tool found that scans the persistent memory layer on disk rather than runtime tracing. Detects injected instructions, obfuscated payloads (zero-width Unicode, base64, homoglyphs, hex), C2-pattern URLs, credential leakage. ratine diff compares snapshots for belief drift.
At a glance
- Type
- Offline scanner — memory poisoning detector
- Tier
- T2
- Created
- 2026-04
- Latest release
- v0.1.0 2026-04-02
- License
- MIT
- GitHub
- 1★ Python
- Pricing
- free / open source (MIT); no paid tier
- Funding
- no public funding data found
Taxonomy
- storage
- file
- retrieval
- extraction-pull
- persistence
- cross-session
- update
- read-only
- unit
- fact
- governance
- auditable
- conflict
- n/a
When to use
Optimised for: memory operation tracing + drift / poisoning detection
Anti-fit: not for use cases that don't run agent workloads in production
Pros & cons
Pros
Memory-specific observability with explicit memory-quality dashboards.
Cons
Newer entrant; small adoption signal; thin documentation.
Claims & capabilities
Free / open source. Auto-detects Mem0 local, AutoGen, custom layouts. Zero deps, zero network calls.
Technical surface
- API surface
- searched not found
- Backend storage
- searched not found
- Deployment
- local CLI — Python; pip install; no server required
- Embedding model
- searched not found
- Multi-tenancy
- searched not found
- MCP
- Yes — built for MCP tool-server ecosystem (sibling tools heddle/loom are MCP runtimes); detects agent memory poisoning across MCP sessions
- A2A
- no data — searched github.com/goweft/ratine README; single-purpose memory-poisoning detector for MCP tool servers, no A2A protocol
- OpenTelemetry
- searched not found
Similar systems
Other memory observability & monitoring in the catalog, ranked by inbound references.
- AgentOps T2
When Mem0 is connected, gains Memory Operation Timeline, Search Analytics, Memory Growth tracking, Error Tracking per memory call. Standalone, records context at each step but doesn't analyse memory quality.
- Galileo (galileo.ai) T1
Treats memory as first-class in multi-agent tracing. Luna-2 SLMs (3B / 8B) scan every interaction for intent drift and belief drift; 20+ checks at sub-200ms latency. Catches when agent A's view of the world splits from teammate B's. OpenTelemetry-compatible.
- Langfuse T1
Memory module reads/writes captured as named spans. Trace Log View concatenates every agent step including memory ops. Agent Graphs (GA 2025) infer graph structure from observation nesting; session-level replay tracks how memory state evolves.
- LangSmith T1
Memory reads, vector DB retrievals, state changes are distinct span types in traces. RAG eval separates retrieval quality (context precision) from generation quality (faithfulness). Dataset versioning guards against eval drift.
Related systems
References (1)
- Mem0 integrates with — Free / open source. Auto-detects Mem0 local, AutoGen, custom layouts.