Cognos Session Memory
This server maintains verified session context across Claude conversations using an epistemic trust scoring formula (C = p·(1−Ue−Ua)) that weighs prediction confidence against uncertainty before injecting context. It exposes two MCP tools: save_session to persist summaries with trust scores and load_session to retrieve verified context above a threshold. Under the hood it runs a FastAPI gateway with a /v1/plan endpoint that extracts structured fields from recent SQLite traces, computes trust scores, and either injects a system prompt or flags low-confidence context for review. Reach for this when you need audit trails of what context was injected into prompts or want to gate retrieval on confidence scores rather than dumping all prior session data indiscriminately.
- GitHub stars
- 1
- tools
- 2
- transport
- STDIO
- env var
- 1
- last updated
- Mar 3, 2026
Tools
2 toolsPublic tool metadata for what this MCP can expose to an agent.
save_sessionTrust-score and persist a session summary
Trust-score and persist a session summary
No parameter schema in public metadata yet.
load_sessionRetrieve last verified context (default threshold: 0.45)
Retrieve last verified context (default threshold: 0.45)
No parameter schema in public metadata yet.
Sponsors
Agent CaffeineSponsoredClose your lid, come back to finished work
SnugSponsoredThe clipboard manager for the AI eraPromote your product hereReach 125k+ visitors monthly
WaxSponsoredBlock anything with supervision, without erasing your iPhonePromote your product hereReach 125k+ visitors monthly
knobsSponsoredSee your site on any phone, any day, in any languageConfiguration
COGNOS_TRACE_DBPath to SQLite trace database (default: data/traces.sqlite3)
Readme
CognOS Session Memory
mcp-name: io.github.base76-research-lab/cognos-session-memory
Verified context injection via epistemic trust scoring for LLMs.
Solves session fragmentation by maintaining verified, high-confidence session context between conversations.
Problem
Large language models suffer from session fragmentation: each new conversation starts without verified context of previous work. This forces repeated explanations, loses decision history, and breaks long-running workflows.
Existing solutions (persistent memory systems, vector retrieval) either:
- Lack trust scores before injection → hallucinations propagate
- Don't audit which context was injected → compliance gaps
- Treat all past information equally → noise overwhelms signal
Solution
A plan-mode gateway that:
- Extracts structured context from 3-5 recent traces
- Scores context quality via CognOS epistemic formula:
C = p · (1 − Ue − Ua) - Injects as system prompt only if
C > threshold - Flags for manual review if
C < threshold - Audits every context injection with trace IDs → EU AI Act compliance
Architecture
Details
- Repository
- github.com/base76-research-lab/cognos-session-memory
- Categories
- AI & LLM Tools, Monitoring & Observability
- Package
cognos-session-memory-mcp- Transport
- STDIO
- Registry
- MCP registry, active
More from base76-research-lab
Related AI & LLM Tools MCP servers
View all
Dingdawg Finance Agentdingdawg-dev/dingdawg-finance-agentFinancial analysis AI. Forecasting, audit prep, SOX/GAAP compliance. Accuracy compounds with use.1starsDingdawg Healthcare Agentdingdawg-dev/dingdawg-healthcare-agentHIPAA-aware healthcare AI. Clinical documentation, compliance checks, audit prep. Free local tier.1stars
Verified Repo Memorycognitivemyriad/mcp-verified-repo-memoryStale-proof repository memory with citations + just-in-time verification + TTL (repo-scoped).1stars
Dingdawg Legal Agentdingdawg/dingdawg-agent-1AI contract review, clause drafting, compliance checklists & legal research.0starsDingdawg Support Agentdingdawg/dingdawg-agent-1AI support agent — ticket classification, response drafting, escalation routing.0stars
Memorysharedmemoryai/mcp-serverPersistent memory for AI agents. Remember, recall, and explore a shared knowledge graph.0stars