Cognos Session Memory
Store session traces and compute CognOS trust scores (C=p·(1-Ue-Ua)) from your agent via local stdio MCP.
Overview
CognOS Session Memory is a MCP server for the Build phase that stores agent session traces and exposes CognOS trust scoring tools over stdio.
What is this MCP server?
- CognOS trust scoring exposed as MCP tools using C=p·(1-Ue-Ua)
- Session trace storage backed by configurable SQLite (COGNOS_TRACE_DB)
- PyPI package cognos-session-memory-mcp with stdio transport
- Version 0.1.0 with local data directory default data/traces.sqlite3
- Research-lab CognOS stack oriented to traceability and trust metrics
- Package and server version 0.1.0
- Default trace database path data/traces.sqlite3 when COGNOS_TRACE_DB unset
- Stdio transport via PyPI identifier cognos-session-memory-mcp
What problem does it solve?
Custom agent workflows lack persisted traces and a consistent trust score your coding agent can query mid-session.
Who is it for?
Solo builders prototyping CognOS-style agents who want local SQLite traces and trust tools inside MCP.
Skip if: Production teams wanting managed APM, multi-tenant trace UI, or wallet or payment scoring unrelated to session trust.
What do I get? / Deliverables
After you install the PyPI MCP package and point COGNOS_TRACE_DB, your agent can store traces and invoke CognOS trust scoring without a separate dashboard.
- Persisted session traces in SQLite for agent runs you define
- MCP-invokable CognOS trust scoring using the documented C=p·(1-Ue-Ua) formulation
- Local, repo-adjacent audit trail for agent prototyping
Recommended MCP Servers
Journey fit
Session memory and trust tooling are part of assembling reliable agent stacks during Build, before you harden monitoring in Operate. Agent-tooling subphase covers MCP servers that persist traces and score confidence for multi-step agent workflows you are wiring up.
How it compares
Local agent memory and trust MCP, not Solana wallet scoring or a hosted design-system API.
Common Questions / FAQ
Who is CognOS Session Memory for?
Indie agent builders and researchers wiring CognOS trust and trace storage into Claude Code or other stdio MCP clients on their own hardware.
When should I use CognOS Session Memory?
Use it in Build agent-tooling while you prototype multi-step agents that need durable session traces and CognOS trust scoring callable as tools.
How do I add CognOS Session Memory to my agent?
Install cognos-session-memory-mcp from PyPI, configure stdio MCP in your client, and optionally set COGNOS_TRACE_DB to your SQLite path.